《No One Did It: Responsibility Laundering, from the Scapegoat to the Algorithm》（第 5 部分 · 第 14–20 章）

《No One Did It: Responsibility Laundering, from the Scapegoat to the Algorithm》（第 5 部分 · 第 14–20 章）

> 作者：李笑来 · 本人在 GitHub 公开的书稿 · [GitHub 仓库](https://github.com/xiaolai/no-one-did-it) · 本馆按 2026-05-31 的版本收录，共 20 章，本页为第 14–20 章。图片以链接形式给出。
> 本篇原文为英文，暂无中文译文，下面是英文原文。

Chapter 11 — Make Responsibility Follow Control

---
chapter_id: 11-make-responsibility-follow-control
chapter_number: 11
part: IV
title: Make Responsibility Follow Control
evidence_grade: A
---

Chapter 11 — Make Responsibility Follow Control

> In those scriptures it has been laid down that kings should stand with the rod of chastisement uplifted in their hands.
>
> — *The Mahabharata*, Book 12, Śānti Parva, Rājadharmānuśāsana Parva, Section CXXI, trans. Kisari Mohan Ganguli (1883–1896)

Three documents, all from the last ten years, all signed by officials acting within their authority. None contains an unlawful sentence. Each leaves a hole the diagnostic of chapter 3 walks into.

The first is the Boeing Deferred Prosecution Agreement of 7 January 2021. The Statement of Facts admits that two named Boeing 737 Technical Pilots made misleading statements to the Federal Aviation Administration's Aircraft Evaluation Group about the Maneuvering Characteristics Augmentation System[¹](#c11-1). The corporation paid approximately $2.5 billion. The agreement filed no indictment against any officer above the Technical Pilot layer. As ch-2 installed, and as ch-5 walked at full depth in the anchor case, the chain that produced the Lion Air and Ethiopian crashes did not climb to the engineering executives who approved MCAS's 2016 authority expansion, did not climb to the Type Rating commercial-differentiator decision, did not climb to the Organization Designation Authorization framework under which Boeing employees signed for the FAA. The agreement closed at the rank the agreement chose to close at. Nothing in its text said which person, by name, bore the consequence if the certification questions failed.

The second document is the 1 August 2002 Office of Legal Counsel memorandum on the meaning of "torture" under 18 U.S.C. § 2340A, signed by Assistant Attorney General Jay S. Bybee and principally drafted by John C. Yoo[²](#c11-2). The memorandum was withdrawn by Acting OLC head Daniel Levin on 30 December 2004[³](#c11-3). As ch-2 installed at the partial-scapegoat layer and as ch-9 returned to at the executive-legal-opinion layer, the OLC's role inside the executive branch is to give the official legal interpretation by which the executive will act. The opinion was authored. The opinion was signed. The opinion was retracted. No person in the OLC chain — author, signatory, supervising Attorney General — was the subject of an SEC-style attestation requirement, a regulator's named-function allocation, or a statutory inquiry with compulsory-evidence power over the drafting process. The opinion is a document. The document had a signature line. The signature line was not, by design, the seam at which any subsequent inquiry could fix individual accountability.

The third document is Executive Order 13328 of 6 February 2004, which chartered the Commission on the Intelligence Capabilities of the United States Regarding Weapons of Mass Destruction — the body we earlier called Robb-Silberman[⁴](#c11-4). Chapter 10 installed the order's load-bearing feature: by its own terms it scoped the Commission's mandate to the intelligence community's product and did not extend it to how policymakers used that product. The order's Section 2 mandate charged the Commission to assess the Intelligence Community's capacity to collect, analyse, and disseminate; it named no authority to examine how policymakers used the resulting product. The order is two pages. It contains no concealment. It does what it says. The scope is the design — by what the order grants, not by what it excludes.

Three designs. Each is a sentence in a corporate plea, a Justice Department memorandum, or an executive order. Each, on its face, contains nothing the diagnostic of chapter 3 names as laundering. The hidden architecture is somewhere else. The hidden architecture is what the document does not contain — a named officer who bears the consequence of the control question failing at the layer the document touches. Each was lawful, candid, and self-explaining; the single thing each was scrupulous enough to leave out was a person.

The fair clue: in each case, the opposite design choice exists in another jurisdiction, or in another statute on the same country's books, or in a different historical moment of the same administration. The variable is not the intent. The variable is the design. What follows is how to read for that variable, and what to do when reading is not enough.

Three documents, no name

The institutional defence of each of the three design absences is on the public record. The defences are not weak. Each has its scholarly literature, its court precedent, its public-policy advocates. We take them at full strength before turning the diagnostic on them.

The defence of the Boeing DPA structure runs through Department of Justice prosecution policy. Corporate criminal liability under *New York Central R.R. v. United States*, 212 U.S. 481 (1909), reaches the corporation through the acts of its agents. The Department's policy under the Principles of Federal Prosecution of Business Organizations conditions corporate plea negotiations on the corporation's cooperation, including the production of evidence against individual wrongdoers[⁵](#c11-5). The defence holds that individual prosecution and corporate plea operate as a single bargaining instrument; conditioning the corporate plea on individual prosecution would, on the defence's account, chill the cooperation by which both are reached. The corporation is the party at the table. The corporation can be punished. The individual prosecution is a downstream question that the cooperation may or may not produce. The fact that no Boeing officer above the Technical Pilot layer was indicted is not, on this account, a design absence. It is the operation of the design.

The defence of the OLC opinion process runs through the constitutional separation of powers. Office of Legal Counsel opinions are internal executive-branch guidance binding on the executive in the absence of judicial review. The OLC's role is to give the executive its best legal advice; the lawyer's role within the executive is the lawyer's role anywhere, namely to advise the client. Attorney–client privilege adheres in part. Imposing an SEC-style attestation requirement on OLC opinions would, on this defence, convert legal advice into a public certification regime that the constitutional design did not contemplate and that would, in effect, give the legal opinion the force of law independent of its policy use. The opinion is a document. The remedy for a bad opinion, on this account, is withdrawal, public criticism, and the appointment of better lawyers. Statutory attestation is the wrong instrument.

The defence of the Robb-Silberman charter scope runs through inquiry-design prudence. An inquiry empowered to examine how policymakers used intelligence would, on this defence, become a political instrument by which one administration adjudicates the previous administration's policy choices. The fenced charter is, on this account, the institutional discipline that distinguishes a technical inquiry from a partisan one. The Commission's mandate, framed by Section 2 of the order around the intelligence community's capacity to collect, analyse, and disseminate, was, in this view, the condition under which any honest examination of the WMD failure could proceed in 2004.

And underlying all three defences sits a fourth, broader defence of the design status quo. It is the libertarian / regulatory-capture line associated with Henry Manne, Frank Easterbrook, and Gary Becker in different forms, and it is the dominant steelman to come: design moves that impose individual responsibility on senior officers risk producing defensive compliance rather than actual accountability, criminalising operational judgment, and deterring qualified individuals from senior roles. We hold this argument open and engage it at full strength below.

Each defence is internally consistent. Each is the public account given by a serious institution of why the design is the design. The diagnostic does not refute any of them at the level on which they operate. The diagnostic asks a different question.

The diagnostic turned forward

The diagnostic of chapter 3 — its eight questions of blame, control, benefit, knowledge, preventability, record-control, cost, and what responsibility would look like if it followed control instead of visibility — has done its work after the harm across the cases we have installed: Bhopal at the chairman layer in chapter 5, family separation through the misdemeanor statute in chapter 6, the Horizon prosecutions through the five-role conflation of the Post Office in chapter 7, the second-term proclamation arc in chapter 8, the AI training-data and deployment grammar in chapter 9, the Iraq WMD case through the National Intelligence Estimate in chapter 10. It walks backward from the visible blamed party. It reaches the named decision-maker who controlled the layer at which the harm could have been stopped.

Chapter 11 turns the same eight questions forward. Run as a design instrument rather than a forensic instrument, each question becomes a specification. *Who would be publicly blamed?* That is the design's exposed surface — the Technical Pilot named in the Boeing DPA's Statement of Facts, before any question of who above him was indicted. *Who would have control?* That is the design's named locus — the Acme Markets president who, by reason of his position, had the authority to direct sanitation at the Baltimore warehouse. *Who would know?* That is the design's attestation point — the chief financial officer signing the § 302 certification under penalty of § 906. *Who could prevent?* That is the design's intervention window — the senior manager allocated, by written Statement of Responsibilities, the function in which the contravention occurred. *Who would control the record?* That is the design's audit boundary — the inquiry chair's section 21 notice, by which Cabinet papers and JIC assessments are compelled into the inquiry's record. Who would bear the cost if the layer fails? That is the design's accountability seam — the named individual whose signature, allocation, or compelled testimony fixes the consequence to a person rather than to a procedural shell.

Four statutes and doctrines, all currently on the books, demonstrate the inversion at four different layers of institutional life.

The first is the Park doctrine of responsible-corporate-officer liability under the Federal Food, Drug, and Cosmetic Act. The doctrine takes its name from *United States v. Park*, 421 U.S. 658 (1975), in which the Supreme Court affirmed the misdemeanor conviction of Acme Markets, Inc. president John R. Park for adulterated-food violations at company warehouses, holding that the FDCA reaches an executive who, by reason of his position in the corporation, has "the responsibility and authority either to prevent in the first instance, or promptly to correct, the violation complained of"[⁶](#c11-6). The doctrine is conventionally described, in corporate-law casebooks, as strict liability under the FDCA's misdemeanor structure, with an "objective impossibility" defense framed around the officer's responsible relation and authority to prevent or correct. The opportunity-to-prevent reading we rely on — the design-rule reading that treats Park as conditioned on the executive's position in the responsibility chain and the executive's opportunity to prevent the violation — is one defensible reading of the Court's text and is associated with the Sepinwall line of scholarship; the casebook line treats Park as strict liability. We use the design-rule reading because it maps Park onto the named-attribution specification we are arguing for. The named officer is the certifying officer of the corporation's compliance at the relevant operating layer.

The second is Sarbanes-Oxley §§ 302 and 906, enacted in the wake of Enron, WorldCom, and the 2002 audit-failure cluster. Section 302, codified at 15 U.S.C. § 7241 (as implemented by SEC Rules 13a-14 and 15d-14), requires the chief executive officer and chief financial officer of each issuer to certify in each periodic report that they have reviewed the report, that based on their knowledge it does not contain material misstatements, that based on their knowledge the financial statements fairly present the issuer's condition, and that they have designed and evaluated the issuer's internal controls[⁷](#c11-7). Section 906, codified at 18 U.S.C. § 1350, attaches a criminal penalty to knowingly false certifications: up to ten years' imprisonment for a knowing violation, twenty for a wilful violation[⁸](#c11-8). The named officers are the CEO and the CFO. The named act is a personal signature attached to a periodic filing. The named state of mind for the criminal penalty is knowing falsity.

The third is the United Kingdom's Senior Managers and Certification Regime. The pattern, in one sentence: every senior function inside a regulated firm must be allocated, in writing, to a single named individual who can be held to it by the regulator. The statutory plumbing follows.

The regime was inserted into the Financial Services and Markets Act 2000 by the Financial Services (Banking Reform) Act 2013. The inserted sections — 59ZA, 59A, and 59B — govern the approval and statements of responsibilities for senior managers. The regime is operationalised through the Financial Conduct Authority's Senior Management Arrangements, Systems and Controls sourcebook. SYSC 4.7 governs the allocation of senior management responsibilities. SYSC 24 governs the allocation of prescribed responsibilities under the regime. SYSC 27 governs the certification-employees layer[⁹](#c11-9).

The regime has three pillars that the design-rule argument must distinguish, because they operate at different layers of the firm. (1) The Senior Managers Regime: an individual exercising a Senior Management Function must be pre-approved by the regulator, must have a written Statement of Responsibilities, and bears the statutory duty under FSMA s.66A(5) to take such steps as a person in their position could reasonably be expected to take to avoid contraventions in their area of responsibility. (2) The Certification Regime, under FSMA ss.63E–63F: the firm itself certifies, annually, that staff in specified non-approved functions are fit and proper; certification is firm-issued, not regulator-approved. (3) The Conduct Rules (FCA COCON): individual-conduct standards applicable to a broader population of staff. The design-rule argument operates specifically at the Senior Managers Regime layer — the SMF + Statement of Responsibilities + s.66A(5) duty — not at the regime as a whole. One further point on burden: the Bank of England and Financial Services Act 2016 removed the regime's original reverse burden of proof. The regulator must now prove that the senior manager failed to take reasonable steps. Under the regime, every Senior Management Function within an authorised firm must be allocated, by a written Statement of Responsibilities, to a named individual. The named officer is whichever individual the firm has allocated the function to. The named record is the Statement of Responsibilities and the firm's responsibilities map.

The fourth is the United Kingdom's Inquiries Act 2005, and within it the compulsory-evidence power granted at section 21. The Act authorises a Minister to establish a statutory public inquiry chaired by a person appointed under section 4[¹⁰](#c11-10). Section 21 empowers the chair, by notice in writing, to require a person to attend, to give evidence, or to produce documents, with criminal penalty for non-compliance under section 35[¹¹](#c11-11). The named officer is the inquiry chair. The named record is whatever the chair compels into the inquiry's record. The named instrument by which the chair extracts cabinet papers, internal correspondence, classified material under appropriate restriction, is a section 21 notice.

Each of these is, in our vocabulary, a *named-attribution design*. Each translates one or more of the diagnostic's eight questions into a specification that operates before the harm rather than after it. The matrix below shows what each design forces an answer to, and what each leaves unmapped.

Read down a column to see what one design does; read across a row to see how the four designs handle the same question.

> *Audio-form rendering of the matrix that follows. Print and screen readers may skip to the table.*
>
> Hold the eight questions against four real designs and look for the empty cells. On who is publicly blamed, each design names someone: the Park doctrine names the responsible officer; the UK Senior Managers and Certification Regime names the senior function-holder; the Inquiries Act 2005, under its section 21, names the inquiry's subjects. (The Sarbanes-Oxley certification column is pending verification and is not asserted here.) On control, the designs differ in reach — the Park doctrine is fenced to the scope of the Food, Drug, and Cosmetic Act, while the Senior Managers regime allocates full senior-function control and a section 21 inquiry takes full subject-matter access. On knowledge and preventability, the strongest instruments tie the named person to an opportunity to prevent: the Park doctrine's responsible-relation standard, the Senior Managers regime's written allocation, the section 21 power to compel disclosure of evidence before findings. On the record, the Senior Managers regime generates Statements of Responsibilities, and section 21 compulsion creates a record where none would otherwise exist. But read down the rows that no design fills. On benefit — who profited — every column is blank: not directly mapped, in any of the four. On the cost-bearer, only a statutory inquiry's findings name who bore the harm. No column is full. Responsibility-following-control is a portfolio, not a single instrument — and the benefit question is unmapped in every design we have.

::: {.figure-embed}
[〔图：The eight diagnostic questions run against four real accountability designs — Park doctrine, SOX, the UK Senior Managers regime, and Inquiries Act §21 — where the gaps are the map of work still undone.〕](https://github.com/xiaolai/no-one-did-it/blob/b5e9c8f820bc04bb1d452d4109318673c08c3945/book/chapters-v6/book/evidence/diagrams/proofs/ch11-design-matrix.png)
:::

No single design covers all eight cells. The italic cells — most notably the Q3 benefit row across all four designs, the Q6 record row at Park, the Q7 cost-bearer row across the corporate-officer designs — are the matrix's most honest content. No single design is sufficient. Responsibility-following-control design is a *portfolio* of moves: Park names the operating officer; SOX names the CEO and CFO at the financial-reporting seam; SMCR allocates each senior function by written statement to one individual; the Inquiries Act compels the evidence by which a chair can name the rest. Each move fills specific cells of the diagnostic matrix. The cells each move leaves unmapped are themselves the map of further design work. Q3 — who benefited — is the cell ch-12 ("Keep the Record") begins to fill, by treating record-discipline as the precondition for tracing benefit flows that occur off the certification timeline. The matrix is not a verdict on the four designs. It is a specification language for the design questions the rest of Part IV continues to work through.

The president of Acme Markets

The Park doctrine is older than nearly every modern laundering case we have treated. Its statutory anchor — the Federal Food, Drug, and Cosmetic Act of 1938 — was enacted thirty-six years before the Bhopal warehouse audit and forty-six years before the Bhopal disaster we installed earlier at the Anderson layer.

The case the doctrine takes its name from is unremarkable on its facts. Acme Markets, Inc. was a national retail food chain. Its president, John R. Park, was charged on a five-count misdemeanor information under the FDCA after the Food and Drug Administration found, in 1971 and 1972, that warehouses operated by Acme were contaminated with rodent infestation that had reached food held for sale. The contamination was real. The FDA's prior warnings to Acme management — including a letter informing Park that the Baltimore warehouse showed evidence of rodent infestation — were on the record. Park's defence at trial was that the company had a written compliance program, that operational control of warehouse sanitation had been delegated to subordinate employees, and that he had no personal involvement in the specific contamination found.

The Supreme Court, in a 6-3 decision authored by Chief Justice Warren E. Burger, affirmed Park's conviction. The relevant passage stands at 421 U.S. 671 to 674. The FDCA, the Court held, imposes "not only a positive duty to seek out and remedy violations when they occur but also, and primarily, a duty to implement measures that will insure that violations will not occur"; it reaches "all persons who, in respect of an act in violation of the Act, share responsibility in the business process out of which the act resulted"; and the requirement that the responsible-relation standard fall within "the range of the agent's authority" preserves the distinction between an officer who could have prevented the violation and one whose role was unrelated to it[¹²](#c11-12).

We read the Court's standard as one of opportunity-to-prevent: a defendant convicted under Park has, by virtue of the position he occupied, the authority and the responsibility to prevent the violation in the first instance or promptly to correct it. (As flagged earlier, the casebook line characterises Park as FDCA strict liability with an objective-impossibility defense; the opportunity-to-prevent framing is our design-rule reading, not the only available one.) The opportunity must be a real opportunity. *Park* itself, on the trial record affirmed by the Supreme Court, turned on the FDA's prior written warning to Park, which the record documents he received[¹³](#c11-13). A subordinate-only structure with no notice to the named officer would not on its face satisfy the standard.

The doctrine has been affirmed and tested at the senior-officer layer in subsequent decades. The *DeCoster* line is the modern test case. In *United States v. DeCoster*, 828 F.3d 626 (8th Cir. 2016), cert. denied, 137 S. Ct. 2160 (2017), the Eighth Circuit affirmed three-month prison sentences imposed on father-and-son Quality Egg executives Austin "Jack" DeCoster and Peter DeCoster under the responsible-corporate-officer doctrine, after the executives pleaded guilty to FDCA misdemeanor counts in connection with the 2010 salmonella outbreak. The panel divided on Eighth-Amendment and due-process objections to incarcerating an RCO defendant who had pleaded guilty without admission of personal knowledge of the violation[¹⁴](#c11-14). *DeCoster* did not narrow Park. *DeCoster* extended Park to actual prison time, against the strongest contemporary constitutional objections, at the senior-officer layer the doctrine was originally designed for. The doctrine exists; the doctrine has been tested; the doctrine has produced prison sentences against named senior officers in the half-century since *Park*; the constitutional objections to incarceration absent personal knowledge are live in the dissents but did not carry.

Park is not, on our reading, a finished design. Park is the *named design move* — codified in statute and judicially expounded — from which the later moves we now walk descend. SOX §§ 302 and 906 are recognisable Park-shaped responses to a different statutory domain. The Senior Managers and Certification Regime in the United Kingdom is recognisable as Park's most extended form, expanded into a written-allocation regime for every senior function in a regulated firm. Each was built knowing the prior move existed. The design tradition has half a century of operating record.

Two signatures, every quarter

The 2002 record-discipline design is the United States' clearest post-*Park* extension to a domain — financial reporting — that *Park* itself did not reach. The Sarbanes-Oxley Act of 2002 was the legislative response to the Enron failure of late 2001 and the WorldCom failure of mid-2002. Among its many provisions, two carry the design-rule argument. They are both responsibility-following-control design moves, and they are both currently in force.

The pattern, in one sentence: two named officers must personally sign each periodic financial filing, and a knowingly false signature carries prison time. The statute does the rest.

Section 302, codified at 15 U.S.C. § 7241, instructs the Securities and Exchange Commission to require, by rule, that the chief executive officer and the chief financial officer of each issuer personally certify each periodic report. The certifying officers must affirm that they have reviewed the report. They must affirm that, based on their knowledge, it does not contain material misstatements or omissions. They must affirm that, based on their knowledge, the financial statements fairly present the issuer's financial condition. They must affirm that they have designed, evaluated, and disclosed the effectiveness of the issuer's internal controls over financial reporting[¹⁵](#c11-15). The SEC's implementing rules — principally Rules 13a-14 and 15d-14 under the Securities Exchange Act — operationalise the certification in the form attached to every Form 10-K and Form 10-Q.

Section 906, codified at 18 U.S.C. § 1350, attaches criminal penalty to knowing or wilful false certification: a knowing violation is punishable by fine and up to ten years' imprisonment; a wilful violation, up to twenty years[¹⁶](#c11-16). The named individual is the certifying officer. The named act is signing. The named state of mind is knowing or wilful falsity.

The criminal-enforcement record under § 906 is small. We do not claim § 906 has produced a wave of CEO prosecutions. We claim something narrower, and the statutory text alone carries the claim. SOX §§ 302 and 906 named the individual required to attest, attached the attestation to a periodic public filing with criminal exposure, and removed the "we delegated to the auditor" alibi from the cleanest layer of corporate reporting.

The narrowness matters because the alternative was not zero accountability. The alternative was diffuse accountability — to the audit committee, to the external auditor, to the corporation in the abstract. Each of those parties existed before SOX. None could be put in the same room as a single named signature on a periodic filing. The certification did not produce a new offence in the criminal-procedure sense; it produced a new *certifying individual* at the seam where the prior diffusion had obscured every individual.

The contrast with the corporate-prosecution design installed in ch-5 sits inside the same country and the same period. The Boeing case under ch-5's anchor was prosecuted by the Department of Justice as a corporate-respondent matter. The 2021 DPA Statement of Facts named the Technical Pilots; the corporate plea did not name an indicted officer above them. The Sarbanes-Oxley regime sits in the same year-range as the design choices that produced the DPA framework, and operates in the same federal architecture. Two designs; same country; opposite specification at the named-officer layer. SOX names the CEO and the CFO and attaches criminal exposure to their signature. The DPA framework names the corporation and attaches financial penalty to its plea. The named individual sits inside one of these designs and does not sit inside the other. The diagnostic predicts where laundering can still operate.

The Park doctrine and the Sarbanes-Oxley §§ 302 and 906 regime together demonstrate that responsibility-following-control design has more than fifty years of operating record at the corporate-officer layer in U.S. statute and federal-court practice. The design is not novel. The design exists. The question we are asking is whether the design will be extended to the layers it has not yet been extended to — and whether the layers at which it operates today are the layers at which the laundering, having displaced, now occurs.

The defensive-compliance objection

The strongest counter to the design-rule argument is not partisan. It does not come from the left or the right. It comes from a body of corporate-law and economic scholarship that has, for a generation, argued that responsibility-following-control statutes of the Park-SOX-SMCR kind produce defensive compliance, not actual accountability.

The argument runs roughly as follows. Imposing personal certification or named-function allocation on senior officers, beyond record-documented knowing misconduct, exposes those officers to liability that they cannot, as a practical matter, reach down through. A CEO of a Fortune 500 company cannot personally inspect every operational layer for which the company's CFO will sign a § 302 certification. A senior manager at a multinational bank cannot, by the architecture of the firm, personally verify every transaction her Statement of Responsibilities makes her the named owner of. The named-attribution design, on this account, therefore does one of three things. It chills the recruitment of senior officers willing to bear the exposure. It generates a permanent compliance overhead — outside counsel reviews, internal certification chains, audit committee theatre — that captures resources without producing accountability. Or it produces a checklist culture in which the auditor signs because the box exists, not because the auditor has verified the underlying conduct. The result is the appearance of responsibility, distributed across senior officers whose practical control of the certified activity is itself a fiction. The argument was advanced in different forms by Henry Manne in commentary on Park, by Frank Easterbrook and Daniel Fischel on responsible-officer liability and the structure of the firm, and by Gary Becker in the broader economic analysis of criminal-deterrence design.

The strong form of this counter adds an empirical observation we must engage. Across the cases we have installed, the most prominent prosecutions of senior individuals — Volkswagen's Oliver Schmidt and James Liang in the United States, Rupert Stadler in Germany — were reached under existing fraud and conspiracy statutes, not under Park-style or SOX-style certification regimes. The Volkswagen executives, on the counter's account, were prosecuted with the criminal-law instruments the U.S. and German systems already possessed; no new responsibility-following-control statute was required. The OLC opinion process that produced the 1 August 2002 Bybee memo was, separately, internal to the executive branch and would not have been altered by a SOX-equivalent attestation regime because internal executive-branch legal advice is not periodic public filing. The 2003-04 Robb-Silberman charter scope was set by executive order, not by inquiry statute; if the political will to set a broader charter had existed, it could have been set without amending statute. The counter therefore concludes: the design rule we argue for would have changed nothing in the cases where it should have worked, and would have produced compliance overhead in the cases where the laundering never approached prosecution.

This is a real argument. It has scholarly weight. It is our dominant steelman.

The response is design-specific and does not contest the counter at the level the counter operates on. It has three parts.

First, the four design moves walked above are not strict-liability moves. Park's standard is opportunity-to-prevent, conditioned on the officer's position and notice. SOX § 906 is conditioned on knowing or wilful falsity. SMCR's named-officer regulatory duty requires the manager to take such steps as a person in the position could reasonably be expected to take. The Inquiries Act § 21 compels evidence; it does not by itself impose individual liability for the conduct revealed. Each design names the conduct that triggers exposure. None operates as universal personal liability for every harm.

Second, the design rule is not "punish the certifier for every operational failure." It is name the certifier as the decision-maker who must choose which layer to inspect, who must record the sub-delegation, and who bears the consequence of failing to inspect the layer at which the harm occurred. Park itself runs on this distinction. The doctrine is recognisable in the slip-opinion text as a structural one: the responsible relation is the relation in which the officer had the authority to designate which subordinate would do the inspecting and to verify that the inspecting had been done. Design happens at the named-attribution layer, not at the strict-liability layer.

Third, the four design moves would not have prevented the cases this book has installed. The narrower and more honest point is that the design moves change the cell of the diagnostic matrix where the laundering must occur. The Volkswagen prosecutions reached the engineering layer; they did not reach the executive layer in the United States, and reached it only partly in Germany. The Bybee memo's authorship survived the OLC's internal withdrawal mechanism because no statutory attestation operated at that layer. The Robb-Silberman charter survived because no statute compelled a broader inquiry charter. Whether the laundering displaced, and to which cell it displaced, is the question the matrix above frames. The libertarian counter is partly correct on whether named-attribution design eliminates laundering. It does not. The question is not regulation vs. no regulation. The question is laundering vs. no laundering. Where the named-attribution layer is filled, the laundering must displace; where the layer is unfilled by design, the laundering does not need to displace at all.

Same country, opposite specification

The defensive-compliance counter is most testable through asymmetry. Four design moves where responsibility-following-control was embedded; four moves where responsibility was fenced. Side by side, one diagnostic question at a time.

It is read horizontally — the design move the laundered system makes, then the design move the responsibility-following-control system makes — at each of four layers we have already named.

##### At the corporate-officer layer.

The DPA-without-individual-indictment move negotiates a corporate plea, names a financial penalty, attaches a compliance monitor, and closes the prosecution without filing an indictment against an executive officer above the operational layer at which the cooperation was reached. The opposite move is the Park doctrine. The FDCA names the responsible officer at the operating layer of the FDA's enforcement domain. The misdemeanor information runs against the named individual whose position carried the authority and the opportunity to prevent the violation. Same general statutory family — both federal regulatory regimes operating against corporations — and opposite specification at the named-officer layer. The DPA closes the chain at the corporate plea. Park names the individual at the seam where the corporate plea would otherwise close.

##### At the financial-reporting layer.

The pre-2002 diffuse-attestation move treated the periodic filing as the product of the company's accounting and audit functions, signed off by the audit committee, certified in form by the external auditor, defended in the abstract by the corporation as the certifying party. The opposite move is SOX § 302. The certification is personal. The signature is by name. The criminal exposure under § 906 attaches to the named officer for knowing or wilful falsity. The corporation continues to exist as a party to the filing. The corporation does not displace the individual. The same form continues to be filed; the named individuals on its certification page are the difference.

##### At the regulated-financial-firm layer.

The pre-SMCR allocation of responsibility within a UK bank or insurer ran through committees, internal control functions, board sub-committees, and the corporate-form abstraction. When the post-2008 enforcement environment asked who had been responsible for what had failed, the answer routinely returned a body, not a person. The opposite move is SMCR. Each senior management function is allocated, by Statement of Responsibilities, to a named individual. The individual bears a regulatory duty to take such steps as the position permits to avoid contraventions in their area. The FCA's enforcement record under the regime — Final Notices issued against named senior managers in retail banks for failures in their allocated function — is the operating record. The corporate form has not vanished; the named function-holder is now visible inside it.

##### At the inquiry-statute layer.

The Robb-Silberman move scopes the inquiry's charter, by executive order, to a sub-component of the question the inquiry purports to examine. The Commission's Section 2 mandate ran to the intelligence community's capacity to collect, analyse, and disseminate; the order named no authority to examine how policymakers used the resulting product. The scope was set by what the order granted, not by what it excluded. The opposite move is the Inquiries Act 2005 § 21. The inquiry's chair can compel the production of documents and the attendance of witnesses, with criminal penalty for non-compliance. The Iraq Inquiry — Chilcot — used the power to obtain Cabinet papers, JIC assessments, and the Blair–Bush correspondence, including the 28 July 2002 minute in which Blair wrote "I will be with you, whatever"[¹⁷](#c11-17). Chapter 10 installed the contrast at the wartime case level. We read the contrast there as the difference between a charter that fenced the alibi and a charter that did not. Read at the design-rule level, the contrast is between an executive-order instrument whose scope is set by the executive being examined and a statutory instrument whose compulsion power is set by Parliament under §21, even where the terms of reference are set by the establishing Minister under §5. Both are executive scope-setting; only the UK regime hard-wires the §21 evidence-compulsion power regardless of the Minister's terms.

Four rows. Each holds the same diagnostic question against the laundered design and the responsibility-following-control design. The right column does not eliminate laundering. The right column changes the cell of the matrix in which the laundering must occur. The Boeing case displaced into the layer above the Technical Pilots. SOX displaced fraud into structures the certification did not reach. SMCR displaced misconduct into the layers below senior-manager allocation. The Iraq Inquiry's compulsion power produced findings the Robb-Silberman charter could not produce, and produced no prosecutions of senior officials it documented. Displacement is the prediction, not the rebuttal. The right-column moves were not built to eliminate laundering. They were built to name the cells in which the laundering could continue to occur. Chapter 3 named the questions as a forensic instrument, run backward from a harm; the same eight questions, run forward, are these design specifications. The questions did not change. The direction did.

Name the person before the design ships

Return to the three documents we opened on. None contained an unlawful sentence; each left a hole where a named officer should have borne the consequence of the control question failing. The design tradition just walked is what fills those holes. The Boeing Deferred Prosecution Agreement closed at the Technical Pilot layer; a Park-shaped specification would have named, on the face of the agreement, the officer whose position carried the authority to prevent the certification failure above that layer. The Bybee–Yoo opinion was authored, signed, and withdrawn with no attestation requirement operating over its drafting; a SOX-shaped specification would have attached a named, prosecutable signature to the layer at which the opinion's legal conclusion was certified as sound. Executive Order 13328 scoped Robb-Silberman away from policymakers' use of the product; an Inquiries-Act-shaped specification would have hard-wired the chair's compulsion power regardless of the order's terms. The holes were not unfillable. They were unfilled by design.

The rule that follows from the matrix and the asymmetry table is portable across statute, charter, contract, and policy document. We state it in protocol form.

> When you propose, sign, or advocate for an institutional design — a statute, a corporate plea structure, an inquiry charter, an oversight body's terms of reference, a contract clause, an internal policy, an AI deployment authorisation — ask whether a named person bears the consequence of the control question failing at each layer the design touches. If no named person bears it, the design is laundering by construction. Add the name to the design before the design is built.

The rule is a test, not a sentiment. It is applied document by document and clause by clause. It is testable against any of the four cells the matrix above identifies: who is named as control-holder, who attests to knowledge, who exercises the prevention authority, who bears the consequence of the layer failing. The rule does not require new statute. The rule requires that the institutional design name the responsibility-following-control specification on the face of its operative text — the way Park names the responsible officer, the way SOX names the CEO and the CFO, the way SMCR names the function-holder, the way the Inquiries Act names the inquiry's chair as the compelling officer.

Our role under the rule is design participant. We sign contracts, vote on referenda, serve on committees, draft charters, allocate roles inside organisations we run or work in. Each of those acts is a design moment. The rule applies at each.

This is our one use of the term *responsibility laundering* here. Our argument is that the term names a recurring institutional method by which power keeps control while moving blame onto a weaker bearer, a procedural shell, or a system without a name. The design rule above is our portable counter-method. It is not the only counter-method we install. It is the first.

Four designs in operation

Four institutional designs in operation. Not hypothetical. Each names the moment of individual accountability the design has produced, the year, the named accountability mechanism's text, and the cost of the escape — the constraint the design places on the powerful actor it binds.

The four, in one line each: **Park** (US, FDCA, 1975 — the responsible officer); **SOX §§ 302/906** (US, 2002 — the CEO and CFO personally certifying, under criminal exposure); **UK SMCR** (2016 — a written Statement of Responsibilities allocating a senior management function to a named individual); and a **section 21 inquiry** under the UK Inquiries Act 2005 (compulsory-evidence power, an inquiry chair compelling the record). Take each in turn.

##### 1. Park doctrine.

*United States v. Park*, 421 U.S. 658 (1975), affirmed by a 6–3 Supreme Court[¹⁸](#c11-18), holds that an officer in a position of "responsibility and authority" to prevent or correct a Federal Food, Drug, and Cosmetic Act violation bears criminal liability under the Act, regardless of personal knowledge — at 421 U.S. 671 to 674. The FDA has continued to use the doctrine across the half-century since *Park*. *United States v. DeCoster*, 828 F.3d 626 (8th Cir. 2016)[¹⁹](#c11-19), tested it at the senior-officer layer: the Eighth Circuit affirmed three-month prison sentences against father-and-son Quality Egg executives Austin "Jack" DeCoster and Peter DeCoster; the Supreme Court denied certiorari the following year. The constraint is what the named executive cannot do — route operational compliance entirely through subordinates and remain criminally untouchable when FDA notice has reached the executive's position.

##### 2. SOX §§ 302 and 906.

Sarbanes-Oxley, enacted 30 July 2002[²⁰](#c11-20), requires the CEO and CFO personally to certify the issuer's internal controls (§ 302), under criminal exposure for knowing falsity (§ 906)[²¹](#c11-21). Every Form 10-K and 10-Q filed by every SEC-reporting issuer since 14 August 2002 bears two named individual signatures with that exposure attached. The criminal-enforcement record under § 906 is modest in volume; the structural effect is broader than the prosecutorial record alone. The argument is that this is, in substantial part, what the named-signature requirement was for — not to produce prosecutions, but to produce signatures that could be prosecuted, attached to the layer at which the prior diffusion had obscured every individual.

##### 3. UK SMCR.

The Senior Managers and Certification Regime, installed under FSMA 2000 ss. 59ZA, 59A, and 59B as inserted by the Banking Reform Act 2013, with FCA Handbook SYSC 4.7, 24, and 27 carrying the operative provisions[²²](#c11-22). It came into force for banks, building societies, and PRA-designated investment firms in 2016 and was extended in 2018–19. The named mechanism is the Statement of Responsibilities — a written allocation of a senior management function to a named individual — together with that manager's regulatory duty to take such steps as a person in that position could reasonably be expected to take to avoid contraventions in the allocated area. The FCA's Final Notices register records enforcement against named senior managers for failures within their function. The regime has not eliminated misconduct in UK financial services. The constraint: under the regime, the named manager cannot point to a committee.

##### 4. Inquiries Act 2005 § 21, contrasted with Executive Order 13328.

The Inquiries Act came into force on 7 June 2005[²³](#c11-23); section 21's compulsory-evidence power is the regime's load-bearing tool. The Iraq Inquiry, chaired by Sir John Chilcot, used the section to obtain Cabinet papers, JIC assessments, and the Blair–Bush correspondence; the report of 6 July 2016 found that the September 2002 dossier had presented JIC assessments with "a degree of certainty that was not justified" and that the legal basis for war had been "far from satisfactory"[²⁴](#c11-24). The contrasting design is Executive Order 13328 of 6 February 2004, which chartered the Robb-Silberman Commission to examine the intelligence community's product on Iraq WMD but, by its own terms, scoped the Commission to intelligence-community capacity and named no authority to examine policymaker use of the intelligence[²⁵](#c11-25). Two designs, two countries, parallel intelligence failure on the same case: the compelling design produced the Cabinet-paper record; the fencing design produced an intelligence-community audit whose conclusions did not extend to the policy use. The constraint under § 21 is the political distance the design imposes on the executive of the day — the chair sets the evidentiary scope, and the executive accepts the compulsion or accepts the public record of its refusal.

Each panel is honest about what the design produced and what the design did not produce. None of the four designs eliminated the laundering this book has installed. Each named the cell of the diagnostic matrix in which the laundering, having displaced, would now be visible. That is the design's named work. The design is not a remedy. The design is the named-attribution specification on which subsequent remedy depends.

The matrix on which we have been operating leaves Q3 — who benefited — across all four moves under-mapped. That cell is what comes next. The design rule names the responsible party. The record discipline is what makes the rule operate.

The design rule reaches our own reading and working life wherever a written document — a board charter, a contract clause, an inquiry's terms of reference, an internal policy — names a function but no person. The recognition trigger is exactly that gap: the scope clause that says "the Committee shall review the firm's internal-control system" with no individual whose written allocation makes the review theirs; the indemnification clause that protects "the Corporation" without naming the officer whose decision triggers it; the "designated person" left as a placeholder. Each is a Park-shaped hole, and the question that surfaces it is one question: at each layer this design touches, who is the named individual whose written allocation makes them the certifying officer for the control question that layer is responsible for? If the answer is a function without a person, a person without a function, or a committee without an allocation, the design absence is the seam. The signature is the seam. Watch the seam.

We are most likely to be positioned as the goat at the layer where our role intersects the design absence. Chapter 6 named the signature as the seam in administrative-law pretext cases. Chapter 11 adds the design-side correlate: do not sign the named-signer line of a function whose written allocation you did not write, whose sub-delegations you were not given, and whose escalation paths route back to the layer that produced the prior failure. Where the written allocation does not exist, ask for it in writing before signing. The written request is itself the design move. The absence of the allocation is the disqualifying condition.

Next we take up the record discipline that makes the design rule operate at the cells of the matrix we have left unmapped.

References

[^384]: United States v. The Boeing Company, Deferred Prosecution Agreement and Statement of Facts, No. 4:21-cr-00005-O (N.D. Tex., filed Jan. 7, 2021), Deferred Prosecution Agreement; Statement of Facts ¶¶ 1-72; Information. [justice.gov](https://www.justice.gov/archives/opa/pr/boeing-charged-737-max-fraud-conspiracy-and-agrees-pay-over-25-billion), [archived snapshot](https://web.archive.org/web/20260526112347/https://www.justice.gov/archives/opa/pr/boeing-charged-737-max-fraud-conspiracy-and-agrees-pay-over-25-billion).
[^385]: Memorandum from Jay S. Bybee, Asst. Atty. Gen., OLC, to Alberto R. Gonzales, Counsel to the President, "Standards of Conduct for Interrogation under 18 U.S.C. §§ 2340–2340A" (Aug. 1, 2002); and DOJ Office of Professional Responsibility, "Report: Investigation into the Office of Legal Counsel's Memoranda Concerning Issues Relating to the Central Intelligence Agency's Use of 'Enhanced Interrogation Techniques' on Suspected Terrorists" (July 29, 2009), Bybee memo Aug. 1, 2002; OPR Report July 29, 2009 (final report released Feb. 19, 2010). [irp.fas.org](https://irp.fas.org/agency/doj/opr-final.pdf).
[^386]: Memorandum from Daniel Levin, Acting Assistant Attorney General, OLC, to James B. Comey, Deputy Attorney General, "Legal Standards Applicable Under 18 U.S.C. §§ 2340-2340A" (Dec. 30, 2004), Levin memorandum of 30 Dec 2004, page 2 (introduction) and footnote 8. [justice.gov](https://www.justice.gov/file/18791/download), [archived snapshot](https://web.archive.org/web/20260526120231/https://www.justice.gov/file/145681/dl?inline=).
[^387]: Executive Order 13328 of February 6, 2004, "Commission on the Intelligence Capabilities of the United States Regarding Weapons of Mass Destruction," 69 Fed. Reg. 6901 (Feb. 11, 2004), 69 Fed. Reg. 6901; Sections 2 and 3 of the Order. [govinfo.gov](https://www.govinfo.gov/content/pkg/WCPD-2004-02-09/pdf/WCPD-2004-02-09-Pg203.pdf), [archived snapshot](https://web.archive.org/web/20260527024057/https://www.govinfo.gov/content/pkg/WCPD-2004-02-09/pdf/WCPD-2004-02-09-Pg203.pdf).
[^388]: U.S. Department of Justice, Justice Manual § 9-28.000, "Principles of Federal Prosecution of Business Organizations", Justice Manual § 9-28.000 (umbrella); § 9-28.700 (cooperation); § 9-28.720 (individuals). [justice.gov](https://www.justice.gov/jm/jm-9-28000-principles-federal-prosecution-business-organizations), [archived snapshot](http://web.archive.org/web/20260425071012/https://www.justice.gov/jm/jm-9-28000-principles-federal-prosecution-business-organizations).
[^389]: United States v. Park, 421 U.S. 658 (1975), 421 U.S. 658, 671-674 (holding); 660-665 (trial-record recitation); decision date 9 June 1975. [law.cornell.edu](https://www.law.cornell.edu/supremecourt/text/421/658), [archived snapshot](http://web.archive.org/web/20250613045039/https://www.law.cornell.edu/supremecourt/text/421/658).
[^390]: Sarbanes-Oxley Act of 2002, Pub. L. No. 107-204, §§ 302 and 906, 116 Stat. 745, codified at 15 U.S.C. § 7241 (§ 302) and 18 U.S.C. § 1350 (§ 906); SEC implementing rules at 17 C.F.R. §§ 240.13a-14, 240.15d-14, Pub. L. 107-204 §§ 302, 906; 15 U.S.C. § 7241; 18 U.S.C. § 1350; 17 C.F.R. §§ 240.13a-14, 240.15d-14. [govinfo.gov](https://www.govinfo.gov/content/pkg/PLAW-107publ204/pdf/PLAW-107publ204.pdf), [archived snapshot](http://web.archive.org/web/20260320115151/https://www.govinfo.gov/content/pkg/PLAW-107publ204/pdf/PLAW-107publ204.pdf).
[^391]: Sarbanes-Oxley Act of 2002, Pub. L. No. 107-204, §§ 302 and 906, 116 Stat. 745, codified at 15 U.S.C. § 7241 (§ 302) and 18 U.S.C. § 1350 (§ 906); SEC implementing rules at 17 C.F.R. §§ 240.13a-14, 240.15d-14, Pub. L. 107-204 §§ 302, 906; 15 U.S.C. § 7241; 18 U.S.C. § 1350; 17 C.F.R. §§ 240.13a-14, 240.15d-14. [govinfo.gov](https://www.govinfo.gov/content/pkg/PLAW-107publ204/pdf/PLAW-107publ204.pdf), [archived snapshot](http://web.archive.org/web/20260320115151/https://www.govinfo.gov/content/pkg/PLAW-107publ204/pdf/PLAW-107publ204.pdf).
[^392]: Financial Services and Markets Act 2000 (UK), sections 59ZA, 59A, 59B, 63E, 63F, 66A, as inserted and amended by the Financial Services (Banking Reform) Act 2013 and subsequent legislation; FCA Handbook SYSC 4.7 (Senior management responsibilities: allocation of responsibilities), SYSC 24 (Senior managers and certification regime: Allocation of prescribed responsibilities), SYSC 27 (Certification regime), FSMA 2000 ss. 59ZA, 59A, 59B, 63E, 63F, 66A; FSBRA 2013; FCA Handbook SYSC 4.7, SYSC 24, SYSC 27; Bank of England and Financial Services Act 2016 s.25. [legislation.gov.uk](https://www.legislation.gov.uk/ukpga/2000/8/contents), [archived snapshot](http://web.archive.org/web/20260515011954/https://www.legislation.gov.uk/ukpga/2000/8/contents).
[^393]: Inquiries Act 2005 (UK), c. 12, sections 1, 4, 5, 21, 35, Inquiries Act 2005 ss. 1, 4, 5, 21(1)-(4), 35; commencement 7 June 2005. [legislation.gov.uk](https://www.legislation.gov.uk/ukpga/2005/12/contents), [archived snapshot](http://web.archive.org/web/20260310233129/https://www.legislation.gov.uk/ukpga/2005/12/contents).
[^394]: Inquiries Act 2005 (UK), c. 12, sections 1, 4, 5, 21, 35, Inquiries Act 2005 ss. 1, 4, 5, 21(1)-(4), 35; commencement 7 June 2005. [legislation.gov.uk](https://www.legislation.gov.uk/ukpga/2005/12/contents), [archived snapshot](http://web.archive.org/web/20260310233129/https://www.legislation.gov.uk/ukpga/2005/12/contents).
[^395]: United States v. Park, 421 U.S. 658 (1975), 421 U.S. 658, 671-674 (holding); 660-665 (trial-record recitation); decision date 9 June 1975. [law.cornell.edu](https://www.law.cornell.edu/supremecourt/text/421/658), [archived snapshot](http://web.archive.org/web/20250613045039/https://www.law.cornell.edu/supremecourt/text/421/658).
[^396]: United States v. Park, 421 U.S. 658 (1975), 421 U.S. 658, 671-674 (holding); 660-665 (trial-record recitation); decision date 9 June 1975. [law.cornell.edu](https://www.law.cornell.edu/supremecourt/text/421/658), [archived snapshot](http://web.archive.org/web/20250613045039/https://www.law.cornell.edu/supremecourt/text/421/658).
[^397]: United States v. DeCoster, 828 F.3d 626 (8th Cir. 2016), cert. denied, 137 S. Ct. 2160 (2017), 828 F.3d 626, 631-639 (majority); 639-654 (dissent by Beam, J.); 137 S. Ct. 2160 (cert. denial). [primary URL no longer resolves as of 2026-05-28; canonical archive] [archived snapshot](http://web.archive.org/web/20241208225602/https://law.justia.com/cases/federal/appellate-courts/ca8/15-1890/15-1890-2016-07-06.html).
[^398]: Sarbanes-Oxley Act of 2002, Pub. L. No. 107-204, §§ 302 and 906, 116 Stat. 745, codified at 15 U.S.C. § 7241 (§ 302) and 18 U.S.C. § 1350 (§ 906); SEC implementing rules at 17 C.F.R. §§ 240.13a-14, 240.15d-14, Pub. L. 107-204 §§ 302, 906; 15 U.S.C. § 7241; 18 U.S.C. § 1350; 17 C.F.R. §§ 240.13a-14, 240.15d-14. [govinfo.gov](https://www.govinfo.gov/content/pkg/PLAW-107publ204/pdf/PLAW-107publ204.pdf), [archived snapshot](http://web.archive.org/web/20260320115151/https://www.govinfo.gov/content/pkg/PLAW-107publ204/pdf/PLAW-107publ204.pdf).
[^399]: Sarbanes-Oxley Act of 2002, Pub. L. No. 107-204, §§ 302 and 906, 116 Stat. 745, codified at 15 U.S.C. § 7241 (§ 302) and 18 U.S.C. § 1350 (§ 906); SEC implementing rules at 17 C.F.R. §§ 240.13a-14, 240.15d-14, Pub. L. 107-204 §§ 302, 906; 15 U.S.C. § 7241; 18 U.S.C. § 1350; 17 C.F.R. §§ 240.13a-14, 240.15d-14. [govinfo.gov](https://www.govinfo.gov/content/pkg/PLAW-107publ204/pdf/PLAW-107publ204.pdf), [archived snapshot](http://web.archive.org/web/20260320115151/https://www.govinfo.gov/content/pkg/PLAW-107publ204/pdf/PLAW-107publ204.pdf).
[^400]: The Report of the Iraq Inquiry — Report of a Committee of Privy Counsellors (Chair: Sir John Chilcot), HC 264, 12 volumes (London: The Stationery Office, 6 July 2016), with declassified annexes including the Blair–Bush correspondence of 28 July 2002, Executive Summary ¶¶ 499 (September Dossier), 510 (JIC certainty caveats), 875-876 (legal basis); Blair-Bush correspondence annexed to Section 3.8 (volume 4). [webarchive.nationalarchives.gov.uk](https://webarchive.nationalarchives.gov.uk/ukgwa/20171123124621/http://www.iraqinquiry.org.uk/), [archived snapshot](https://web.archive.org/web/20260526113028/https://webarchive.nationalarchives.gov.uk/ukgwa/20171123124621/http://www.iraqinquiry.org.uk/).
[^401]: United States v. Park, 421 U.S. 658 (1975), 421 U.S. 658, 671-674 (holding); 660-665 (trial-record recitation); decision date 9 June 1975. [law.cornell.edu](https://www.law.cornell.edu/supremecourt/text/421/658), [archived snapshot](http://web.archive.org/web/20250613045039/https://www.law.cornell.edu/supremecourt/text/421/658).
[^402]: United States v. DeCoster, 828 F.3d 626 (8th Cir. 2016), cert. denied, 137 S. Ct. 2160 (2017), 828 F.3d 626, 631-639 (majority); 639-654 (dissent by Beam, J.); 137 S. Ct. 2160 (cert. denial). [primary URL no longer resolves as of 2026-05-28; canonical archive] [archived snapshot](http://web.archive.org/web/20241208225602/https://law.justia.com/cases/federal/appellate-courts/ca8/15-1890/15-1890-2016-07-06.html).
[^403]: Sarbanes-Oxley Act of 2002, Pub. L. No. 107-204, §§ 302 and 906, 116 Stat. 745, codified at 15 U.S.C. § 7241 (§ 302) and 18 U.S.C. § 1350 (§ 906); SEC implementing rules at 17 C.F.R. §§ 240.13a-14, 240.15d-14, Pub. L. 107-204 §§ 302, 906; 15 U.S.C. § 7241; 18 U.S.C. § 1350; 17 C.F.R. §§ 240.13a-14, 240.15d-14. [govinfo.gov](https://www.govinfo.gov/content/pkg/PLAW-107publ204/pdf/PLAW-107publ204.pdf), [archived snapshot](http://web.archive.org/web/20260320115151/https://www.govinfo.gov/content/pkg/PLAW-107publ204/pdf/PLAW-107publ204.pdf).
[^404]: Sarbanes-Oxley Act of 2002, Pub. L. No. 107-204, §§ 302 and 906, 116 Stat. 745, codified at 15 U.S.C. § 7241 (§ 302) and 18 U.S.C. § 1350 (§ 906); SEC implementing rules at 17 C.F.R. §§ 240.13a-14, 240.15d-14, Pub. L. 107-204 §§ 302, 906; 15 U.S.C. § 7241; 18 U.S.C. § 1350; 17 C.F.R. §§ 240.13a-14, 240.15d-14. [govinfo.gov](https://www.govinfo.gov/content/pkg/PLAW-107publ204/pdf/PLAW-107publ204.pdf), [archived snapshot](http://web.archive.org/web/20260320115151/https://www.govinfo.gov/content/pkg/PLAW-107publ204/pdf/PLAW-107publ204.pdf).
[^405]: Financial Services and Markets Act 2000 (UK), sections 59ZA, 59A, 59B, 63E, 63F, 66A, as inserted and amended by the Financial Services (Banking Reform) Act 2013 and subsequent legislation; FCA Handbook SYSC 4.7 (Senior management responsibilities: allocation of responsibilities), SYSC 24 (Senior managers and certification regime: Allocation of prescribed responsibilities), SYSC 27 (Certification regime), FSMA 2000 ss. 59ZA, 59A, 59B, 63E, 63F, 66A; FSBRA 2013; FCA Handbook SYSC 4.7, SYSC 24, SYSC 27; Bank of England and Financial Services Act 2016 s.25. [legislation.gov.uk](https://www.legislation.gov.uk/ukpga/2000/8/contents), [archived snapshot](http://web.archive.org/web/20260515011954/https://www.legislation.gov.uk/ukpga/2000/8/contents).
[^406]: Inquiries Act 2005 (UK), c. 12, sections 1, 4, 5, 21, 35, Inquiries Act 2005 ss. 1, 4, 5, 21(1)-(4), 35; commencement 7 June 2005. [legislation.gov.uk](https://www.legislation.gov.uk/ukpga/2005/12/contents), [archived snapshot](http://web.archive.org/web/20260310233129/https://www.legislation.gov.uk/ukpga/2005/12/contents).
[^407]: The Report of the Iraq Inquiry — Report of a Committee of Privy Counsellors (Chair: Sir John Chilcot), HC 264, 12 volumes (London: The Stationery Office, 6 July 2016), with declassified annexes including the Blair–Bush correspondence of 28 July 2002, Executive Summary ¶¶ 499 (September Dossier), 510 (JIC certainty caveats), 875-876 (legal basis); Blair-Bush correspondence annexed to Section 3.8 (volume 4). [webarchive.nationalarchives.gov.uk](https://webarchive.nationalarchives.gov.uk/ukgwa/20171123124621/http://www.iraqinquiry.org.uk/), [archived snapshot](https://web.archive.org/web/20260526113028/https://webarchive.nationalarchives.gov.uk/ukgwa/20171123124621/http://www.iraqinquiry.org.uk/).
[^408]: Executive Order 13328 of February 6, 2004, "Commission on the Intelligence Capabilities of the United States Regarding Weapons of Mass Destruction," 69 Fed. Reg. 6901 (Feb. 11, 2004), 69 Fed. Reg. 6901; Sections 2 and 3 of the Order. [govinfo.gov](https://www.govinfo.gov/content/pkg/WCPD-2004-02-09/pdf/WCPD-2004-02-09-Pg203.pdf), [archived snapshot](https://web.archive.org/web/20260527024057/https://www.govinfo.gov/content/pkg/WCPD-2004-02-09/pdf/WCPD-2004-02-09-Pg203.pdf).

Chapter 12 — Keep the Record

---
chapter_id: 12-keep-the-record
chapter_number: 12
part: IV
title: Keep the Record
---

Chapter 12 — Keep the Record

> It would be better to be without the Book of History than to give entire credit to it.
>
> — Mencius, 7B.3, trans. James Legge

Across nights in October and November 1969, two men photocopied a study they were not authorised to copy. The copy sat in a RAND Corporation safe in Santa Monica. The photocopier sat in an advertising agency on Melrose Avenue in Hollywood. The two men moved the pages between the two buildings after hours.

The study had been ordered by the Secretary of Defense, Robert S. McNamara, in June 1967. Its title, when complete, was *United States–Vietnam Relations, 1945–1967: A Study Prepared by the Department of Defense*. It ran to forty-seven volumes and roughly seven thousand pages. It was classified top secret. Daniel Ellsberg, a former Marine company commander, RAND Corporation analyst, and consultant to the study itself, had access to the copy held at RAND's Santa Monica offices. Anthony Russo, also of RAND, helped him operate the photocopier across successive nights in the offices of Linda Sinay, who ran an advertising agency on Melrose Avenue in Hollywood and made the office's copier available after business hours. The photocopying took months.

Eighteen months later, on 13 June 1971, *The New York Times* began publication of what it called the Pentagon Papers, the first instalment of a planned series drawn from the photocopied study. The Nixon administration moved within days for prior restraint. A district court enjoined further publication; the Second Circuit stayed the case for argument; *The Washington Post* picked up where the *Times* had left off. On 30 June 1971, the Supreme Court of the United States, sitting en banc and ruling per curiam, held in *New York Times Co. v. United States*, 403 U.S. 713 (1971), that the government had not met the heavy burden required for prior restraint of the press.[¹](#c12-1) Justice Hugo Black, concurring, wrote that the founders gave the free press the protection it needed "so that it could bare the secrets of government and inform the people."[²](#c12-2)

The Pentagon Papers were not the only record that survived destruction or suppression across the period this book has documented. They were not the most consequential. They were not even the only one inside the Nixon archive. What they were was visible. A study existed because a Secretary of Defense had ordered it written. A copy existed outside the originating institution because a man with access had photocopied it across nights. A newspaper published because its editors decided to publish. A court ruled because the litigation reached it within seventeen days.

The fair clue is in the verbs. The study existed. The copy existed. The newspaper published. The court ruled. Each verb has a documented human author. We return to the verbs at the close.

National security, candour, routine

For most of the period we have documented, institutions have offered the same three defences when records they would prefer to suppress, classify, or destroy have come under pressure to surface. The defences are not always wrong. Each protects a real institutional interest. Taken together, they describe the procedural envelope inside which record-control laundering operates.

The first defence is national security. Classification of information that would, if disclosed, prejudice the security of the state is a standing feature of every modern executive branch. The Pentagon Papers were classified top secret in 1969. The Iran-Contra communications channels were classified in 1986. The intelligence assessments underlying the 2003 Iraq invasion were classified across the period chapter 10 walked. In early 2025, the Department of Justice records-destruction events — the J6 case-tracking dashboard taken offline, the named-prosecutor terminations chapter 8 documented — were defended in administration communications in similar terms. The national-security defence does real work. It also operates, in the cases this book has documented, as the standing rationale. Classification becomes the institutional mechanism by which records most embarrassing to the classifier remain unavailable to the inquiry, the court, or the citizen.

The second defence is deliberative-process privilege. The principle that internal advice to decision-makers needs space to be candid is older than the modern administrative state. American case law from *United States v. Reynolds*, 345 U.S. 1 (1953), through the long *Judicial Watch v. NARA* litigation has read the doctrine generously. The British equivalent — the "thirty-year rule" before its 2010 amendment, the public-interest balancing test under the Freedom of Information Act 2000, the Cabinet's standing claim that papers shown to a particular Prime Minister are protected from successor disclosure — operates on the same logic. Raw thinking, the argument goes, needs space. Officials who write their honest assessments knowing the assessments will be public within a year will not write honest assessments. The official record will then be worse, not better. The deliberative-process defence does real work too. It also operates as the standing rationale by which the documents that would record what a senior official was told, on what date, against what advice, are protected from the inquiry that would use them.

The third defence is routine. Records are destroyed all the time. Every institution operates under retention schedules — paper records discarded after seven years, emails after ninety days, transitory communications after thirty. The Records Disposal Acts and their successors across multiple jurisdictions exist precisely because records cannot be kept forever. When a record is missing, the first institutional account is rarely "we destroyed it deliberately." It is "it was destroyed under the standard schedule." The defence is procedurally correct, often. It is also the rationale that covered Lt. Col. Oliver North's shredding of National Security Council documents across November 1986. It is the rationale under which the Department of Government Efficiency moved, in early 2025, to characterise encrypted-messaging traffic as not constituting federal records (*CREW v. U.S. DOGE Service*, No. 1:25-cv-00511, D.D.C., chapter 8). It is the rationale under which OpenAI's standard log-retention schedule operated to delete ChatGPT conversation logs that the Southern District of New York preservation order in *NYT v. Microsoft / OpenAI* (S.D.N.Y., MDL, chapter 9) then required to be preserved.

Each defence, taken alone, protects something the alternative would not protect. National-security classification keeps live operational information from adversaries. Deliberative-process privilege protects the candour of internal advice. Routine destruction prevents the accumulation of paper into immovable mountains. The three defences also describe, when stacked, the procedural shell inside which the records most damaging to the actor with control of the records can be made unreachable by inquiry, court, or successor.

Three portable rules

Chapter 7 installed the institutional form of record-control laundering. Horizon. The Post Office held five roles against the sub-postmasters it prosecuted — deployer, audit-trail controller, complainant, prosecutor, and disclosure controller — and the engineering record the prosecutions depended on was held by the institution making the accusation. The anti-laundering rule there was: the record is first. The eight-question diagnostic only operates on cases where the record is accessible and was not produced by the party being diagnosed. Where the record is captured, the diagnostic returns the captor's preferred answer.

Chapter 11 added the design rule. Named attribution. A decision belongs to a named person on a documented date against a documented alternative. Anonymous structural decisions are the form responsibility takes when it has been laundered.

Chapter 12 is the layer that makes both prior rules operate. A named decider whose decisions leave no documented trace has named responsibility on paper only. A record-control architecture that allows silent destruction has chosen to be defeatable. Record discipline is not bureaucratic overhead. It is the precondition of every remedy chapters 1 through 10 documented. Every counter-case we have installed turned on a record that survived an institution's interest in suppressing, classifying, destroying, or sequestering it. The mechanisms of survival are nameable. We now abstract them into three portable rules.

##### Rule one. Write it down.

A decision that was not written did not exist, for the purposes of any future inquiry, court, regulator, journalist, or successor. The first record-discipline rule is simply this: when control is exercised, when a decision is made, when an exception is approved, the decision and its maker enter the record on the same day, in writing, with the named individual's signature. The rule answers chapter 11's question six (record) at the design layer.

Two worked examples from prior material. The OLC opinion-shop documented at chapter 2 (Jay Bybee, Aug. 1, 2002; John Yoo across the same period) and again at chapter 8 (the Alien Enemies Act memoranda of March 2025; the January 6 termination memoranda of January and February 2025) operates because Office of Legal Counsel opinions are not subject to Sarbanes–Oxley-style certification by their author or their reviewing official. The opinion is signed, but the signature does not carry the certification weight that a corporate financial filing's signature carries. The opinion can be replaced, withdrawn, reframed by a successor administration without a contemporaneous record of why. The rule applies: an opinion that operates as legal authorisation for executive action enters the record with the named author, the date, the named reviewing official, the operative conclusion, and the alternatives considered, in a form discoverable by successor administrations and statutory inquiries. The rule does not yet exist as statute. It is the next design move we advocate.[³](#c12-3)

The second worked example is the Horizon five-role conflation itself (chapter 7). The Post Office's investigative decisions about individual sub-postmasters — the choice to investigate one branch, the choice to prosecute one sub-postmaster, the choice to deny disclosure of the Known Error Log to a particular defence — were made by named officials inside the Post Office's security and legal functions. Many of those decisions, as the Post Office Horizon IT Inquiry transcript record is reconstructing, were not contemporaneously documented in a form that would have survived institutional turnover. Chain-of-command knowledge inside the Post Office is being reconstructed in 2024 and 2025 from depositions, recovered emails, and the surviving fragments of internal communications. The reconstruction is itself the cost of the rule's absence. Had the contemporaneous decisions about each investigation entered an externally durable record at the point of decision, the reconstruction would not be necessary.

##### Rule two. Timestamp and chain-of-custody.

A record that cannot be timestamped to a moment and traced to a custody chain is evidence of nothing. The second rule is technical: contemporaneous notes timestamped at creation; chain-of-custody logging when records move between custodians; write-once or hash-chained storage where the stakes warrant it. The rule answers the audit-trail gap chapter 7 named at Horizon (the layered record-hardening architecture) and chapter 2 named at Therac-25 (the single memory-mapped error-code counter that was not an audit trail).

The legal-evidence ancestor is named. In the United States, Federal Rule of Evidence 901 requires that evidence be authenticated — its proponent must produce evidence "sufficient to support a finding that the item is what the proponent claims it is."[⁴](#c12-4) Rule 901(b) lists illustrative methods of authentication. Among them: testimony of a witness with knowledge. Comparison by expert. Distinctive characteristics. System or process producing accurate results. The chain-of-custody requirement is the operational form Rule 901 takes when the evidence is a record produced by a system whose output is offered for the truth of what it states. In the United Kingdom, the common-law chain-of-custody doctrine performs the parallel work. It has been supplemented since 1999 by the post-PACE 1984 §69 repeal, under which computer-produced evidence is presumed reliable until rebutted — the presumption chapter 7 named as the second of Horizon's five layers of record-hardening.

The rule we advocate is structural. Chain-of-custody is a multi-party requirement, not a single-party logging requirement. Horizon's Known Error Log existed and was timestamped, but the chain-of-custody between Fujitsu (the engineering custodian) and the Post Office (the deploying custodian, the prosecutor, and the disclosure controller) was controlled by one institution's contract. Where a record that may be used as evidence against a person sits inside the custody chain of the institution making the accusation, the chain is not a chain. It is a closed loop. The rule extends Rule 901's ancestor into the cases we have documented: where the record's accuracy depends on a system or process, the chain-of-custody must include at least one custodian outside the accuser's reporting line.

##### Rule three. Make destruction visible.

The third rule is the hardest of the three, and the one our documented record-failures most often turn on. A record-keeping regime that allows silent destruction has chosen to be defeatable. The rule: destruction is itself a record. The destruction order, the destruction date, the destruction authoriser, and the destruction inventory enter the record before the underlying record is destroyed.

The discipline is borrowed from the litigation-hold structure already operating in American and English civil practice. In the United States, Federal Rule of Civil Procedure 37(e) governs sanctions for failure to preserve electronically stored information that ought to have been preserved in anticipation of litigation. The 2015 amendment narrowed the rule from common-law negligence-spoliation toward a graduated regime: lesser measures for prejudice, the harshest measures only on a finding of intent to deprive.[⁵](#c12-5) The rule operates at the trigger of litigation. We advocate extending its structural logic to administrative records beyond active-litigation triggers — to the records the Federal Records Act, 44 U.S.C. ch. 31, §§ 3101 *et seq.*, and the Presidential Records Act, 44 U.S.C. §§ 2201 *et seq.*, already nominally cover but which the case law chapter 8 documented (the *Armstrong* and *Judicial Watch* arcs) has historically declined to enforce in real time.[⁶](#c12-6)

The worked example is in real time. The Department of Government Efficiency records litigation, *CREW v. U.S. DOGE Service*, No. 1:25-cv-00511 (D.D.C., chapter 8), is testing the rule as we go to manuscript. The administration's position has been that encrypted-messaging traffic does not constitute federal records, that the new service is not an "agency" subject to the FOIA, and that records of personnel embedded inside other agencies belong to those agencies. Judge Christopher R. Cooper entered a preliminary injunction in March 2025 directing preservation of DOGE records.[⁷](#c12-7) The litigation has not produced a final ruling on the agency-status question as of May 2026. What it has produced is a contemporaneous record of which records were proposed for destruction, by which actors, against which preservation arguments. The destruction has been made visible by the litigation itself. The rule, generalised, asks why the visibility had to come from an external lawsuit rather than from the institution's own destruction protocol.

Record discipline is not paperwork; it is the layer at which named attribution becomes operative or decays.

A photocopier on Melrose Avenue

The historical anchor for record discipline is the Pentagon Papers. We walk them once more, compactly, because they teach the three rules in operation by accident rather than design.

The McNamara study existed because McNamara ordered it written. A Secretary of Defense made a decision in June 1967 and signed the order. That is rule one — write it down — in operation, even though the writing was protected by classification, even though the writing was intended as institutional analysis rather than as preservation. The decision generated a record. The record survived because it was generated.

The McNamara study survived suppression because copies existed outside the originating institution. Daniel Ellsberg's photocopying across late 1969 produced a chain of physical objects — boxes of photocopied pages — held by Ellsberg and Russo personally, then by Senator Mike Gravel of Alaska, then in the institutional archives of *The New York Times* and *The Washington Post*. The chain of custody, after Ellsberg's first photocopies, included custodians outside the originating institution's reporting line. The institution that had classified the study could no longer recall every copy. That is rule two — chain-of-custody distributed across custodians — in operation, even though the distribution was unauthorised at its origin, even though the law would later have to determine whether the unauthorised distribution was protected by the First Amendment.

The publication decision made suppression visible. When *The New York Times* began publishing on 13 June 1971, the question whether to suppress the study could no longer be answered inside the executive branch's record-control envelope. The Nixon administration moved for prior restraint. The motion was filed in court. The court ruled. The Supreme Court ruled. The ruling, *NYT v. United States* per curiam against the restraint, entered the United States Reports. That is rule three — make destruction or suppression visible — in operation, even though "destruction" in this case took the form of attempted suppression rather than physical shredding. The attempt entered the record. The record could be tested.

One careful observation belongs here. We are not claiming that Ellsberg's mechanism is universalisable. Ellsberg faced felony charges under the Espionage Act of 1917. The charges were dismissed in May 1973 by Judge William Matthew Byrne Jr. of the Central District of California on grounds of government misconduct — Nixon's White House had ordered the burglary of Ellsberg's psychiatrist Dr. Lewis Fielding's office in September 1971, and the dismissal followed from that misconduct rather than from a ruling on the underlying conduct. The cost of survival was personal. The mechanism that produced survival in 1969–1971 is not a mechanism any institutional design can require. What we take from the Pentagon Papers is not the mechanism. It is the *features* the mechanism happened to satisfy. The study existed in writing. Copies existed outside one custodian's control. The suppression attempt was forced into a venue where it could be tested. Three features. Three rules. Our three rules are abstracted from the features.

Backup tapes the destroyers forgot

Independent Counsel Lawrence E. Walsh, appointed under the Ethics in Government Act of 1978, Title VI (28 U.S.C. §§ 591–599), as in force in December 1986, faced from the first month a record-destruction problem his predecessors had not faced.[⁸](#c12-8)

In November 1986, before Walsh's appointment, the Iran-Contra paper record was reduced inside the office of Lt. Col. Oliver L. North of the National Security Council. On the evening of November 21, North's secretary, Fawn Hall, fed an office shredder roughly a foot and a half of documents, in piles of twelve to eighteen pages, over thirty to sixty minutes, drawn from a five-drawer safe in North's office. On North's instruction, she replaced some of the original documents with altered versions. On November 25, she carried about sixteen pages out of the building concealed in her boots and the back of her skirt.[⁹](#c12-9) Across November 22 through 29, North deleted 736 electronic messages from the NSC's PROFS system, and Vice Admiral John M. Poindexter, then National Security Advisor, deleted 5,012.[¹⁰](#c12-10) The Tower Commission, reporting in February 1987, had documented that destruction but had not been able to reconstruct what had been destroyed.[¹¹](#c12-11) The official record, as it stood when Walsh's office began its work, contained more documented evidence of destruction than evidence of what had been destroyed.

What investigators found, across the next several years, was that the destruction had been partial in a way the destroyers had not anticipated. The National Security Council in 1985 and 1986 used an electronic-mail system called PROFS — Professional Office System — running on IBM mainframes operated by a separate executive-branch communications service. PROFS messages travelled across that service's infrastructure. They were also archived, automatically, on backup tapes operated by yet another office. The shredding at the NSC had destroyed paper. The backup tapes existed. North and Poindexter had deleted 5,748 PROFS messages between them; a separate office, on its standard schedule, had been copying the same traffic to tape the whole time.

The recovery sequence has three documented steps, and the operative escape sits at the first. On November 28, 1986, career staff at the White House Communications Agency ordered the backup tapes preserved rather than recycled under the standard schedule.[¹²](#c12-12) The Tower Commission, reporting on February 26, 1987, used the preserved tapes and reprinted hundreds of PROFS notes in its report.[¹³](#c12-13) Walsh's forensic team, operating across 1987 through 1992, built on that recovery — extracting further volumes of contemporaneous communication from the same backup architecture, including communications that established what the paper record alone could not have established. The chain-of-custody around the tapes exceeded the destroyers' awareness. The chain-of-custody around the *preservation decision*, in turn, exceeded the destroyers' authority: a career-staff action at the communications service intervened before the institutional incentive to recycle could overwhelm the recoverable record.

The Walsh Final Report of August 4, 1993, *Final Report of the Independent Counsel for Iran/Contra Matters*, runs to three volumes. It documents the reconstruction in detail.[¹⁴](#c12-14) The report's most cited single passage concerns President George H. W. Bush's pardons of December 24, 1992, of six Iran-Contra figures, including former Secretary of Defense Caspar Weinberger before his trial. In a contemporaneous statement issued the same day and later incorporated into the Final Report, Walsh declared: "the Iran-contra cover-up, which has continued for more than six years, has now been completed."[¹⁵](#c12-15) The cover-up was, in Walsh's reading, the destruction architecture. What had survived had survived because the destroyers did not know about the backups. Rule three — make destruction visible — applied in the negative: the destruction was made visible by the survival of records the destroyers thought they had destroyed, not by any contemporaneous logging of the destruction itself.

One short paragraph we cannot avoid. What we are writing is itself a record. Every claim in chapters 1 through 11 carries an evidence grade — A for high-confidence public record, B for credible secondary, C for live and contested, D excluded. Every quotation is verbatim and attributed to a single named source, date, and venue per the project's quote-integrity rule. Every citation in the source ledger names the claim, the source, the page or paragraph or timestamp, and the verifier. Writing about responsibility laundering while cutting corners on our own record would launder our own authority. Source-ledger discipline is the same rule at the smallest scale.

The Walsh reconstruction's cost was substantial. The investigation ran for nearly seven years. Walsh's team had to compel disclosure across multiple executive-branch components against contemporaneous political pressure. The Bush 41 pardons of December 24, 1992 — Proclamation 6518 — closed the criminal-accountability track before Weinberger's trial could proceed. (Chapter 8 walked the pardons at the cost-of-escape layer.) The reconstruction produced a documentary record sufficient to support the report's findings. It did not produce the executive-layer criminal accountability the investigation had been authorised to pursue.

When the channel goes dark

Writing the rules and stopping there would skip the strongest argument against them. The argument is real, and we take it at full strength.

The argument is that aggressive record-keeping requirements drive decision-making out of the recorded channels and into the unrecorded ones. The strongest published form of the argument lives in the executive-branch-procedure literature — Adrian Vermeule on the costs of full proceduralism, Cass Sunstein and others on the limits of disclosure as a regulatory tool — and in the practical-administration literature on the deliberative-process privilege. The argument has several components, and they reinforce each other.

First, deliberative-process privilege exists because raw thinking needs space. Officials who write their honest assessments knowing the assessments will be public within a year will not write honest assessments. They will write defensible ones. The official record will then contain less of what the official actually thought and more of what the official wanted the future inquiry to find. Aggressive write-it-down requirements produce a sanitised record, not a more honest one.

Second, the cost of a fully-recorded executive branch is paralysis on consequential matters. Decisions that would benefit from rapid iteration — military planning, diplomatic engagement, regulatory triage during a crisis — slow down when every preliminary thought enters the record. Decision-makers begin to weight, against the merits of any preliminary position, the cost of having that position preserved against them. The merits lose.

Third — and this is the argument's strongest empirical move — when the formal record becomes harder to use, the actual decision-making displaces. Phone calls replace emails. Signal messages replace memoranda. Off-the-books meetings replace recorded ones. The displacement is documented across decades. The Vermeule literature on the costs of FOIA expansion in the 1970s and 1980s describes it. The Department of Justice's own internal documents on Signal use across the second-term period chapter 8 walked are testifying, in real-time docket disclosures, to a contemporary version of it. The displacement is real. A regime that records everything ends up recording the sanitised performance, and the real decisions happen in channels the regime cannot reach.

The argument is not abstract. The displacement-to-unrecorded-channels move has been documented as a strategic choice in active litigation. *American Oversight v. DOGE, Musk, Davis, Gleason, and Rubio* (D.D.C., chapter 8) alleged use of Signal and Google Docs to circumvent the Federal Records Act. The litigation is contested as of May 2026. Our point is not which side prevails. The point is that the displacement is the dominant counter to our third rule, and we cannot answer the counter by pretending it does not exist.

Our response will not pretend the counter dissolves. Rule one in its strong form — when control is exercised, when a decision is made, when an exception is approved, the decision and its maker enter the record on the same day — does carry the displacement cost Vermeule names. The cost is real. Some deliberation that would have happened on paper will move into channels that resist preservation. Some preliminary positions that would have entered the record will be withheld. The sanitised-record critique is not imaginary. We accept the cost and argue that the trade is worth it on the asymmetry table's evidence. The surviving column contains cases — Pentagon Papers, Iran-Contra reconstruction, Iraq Inquiry, Horizon disclosure, Hofeller files — in which the existence of a written, signed, dated decision was the precondition of later accountability. The suppressed column contains cases — Bybee/Yoo OLC, Boeing pre-DPA MCAS, second-term DOJ termination memoranda — in which the absence of write-it-down discipline foreclosed the reconstruction. The displacement counter says the recorded layer will be sanitised. The asymmetry table says the unrecorded layer alone produces nothing for an inquiry to use. A sanitised record beats no record. The trade is the diagnostic move we advocate, not a side-step.

Rules two and three are narrower by construction. Rule two applies to records that have entered some custody chain, not to ephemeral spoken exchanges. Rule three applies to acts of destruction or non-preservation of records that already exist, not to records that were never made. The displacement counter bites hardest at rule one. We have named the bite and accepted it.

What the rules do is shift the cell where the laundering must occur. This is the same structural argument chapter 11 made about its diagnostic matrix. Laundering, under the rules, displaces into the unrecorded channels. But the unrecorded channels have their own structural limits. Phone calls are remembered by their participants and become evidence in later testimony. Signal messages have been the subject of preservation orders in multiple jurisdictions — the *NYT v. Microsoft / OpenAI* preservation order in the Southern District of New York, entered in November 2025 and modified by subsequent orders, compels preservation of model output logs and training-data records (chapter 9). Off-the-books meetings are remembered by their participants. Each unrecorded channel is harder to reach than the recorded channel. None is structurally outside the diagnostic.

Being honest matters. The rules do not eliminate laundering. They shift it. A system with the three rules is better-positioned for later accountability than a system without them. The asymmetry table that follows is our empirical support for that comparative claim.

The subordinate counter — that the United States already has the Federal Records Act, the Presidential Records Act, and the FOIA; that the United Kingdom has the Public Records Acts and the FOI Act 2000; and that the failures we have documented are enforcement failures rather than statutory gaps — we acknowledge next. We advocate enforcement architecture, not new statutory text.

Ten record fights, two columns

Place ten record fights from chapters 1 through 10 in two columns. The pattern in the table is our argument, and the argument is empirical.

A note for listeners: what follows renders as a table in the printed edition. Read aloud, it groups into two columns. The surviving column holds five records. The Pentagon Papers in 1971, surviving under rules one and two — written by McNamara's order, copies outside the originating institution. Iran-Contra across 1986 to 1993, surviving partially under rule two — the PROFS backup chain exceeded the destroyers' awareness. The Iraq Inquiry Cabinet papers across 2009 to 2016, surviving under rule three — statutory compulsion under Inquiries Act 2005 section twenty-one. The Hofeller files in 2018 and 2019, surviving under rule two — posthumous custody change. The Bates-era Horizon disclosure across 2017 to 2019, surviving under rules two and three — the Group Litigation Order compelled custody disclosure. The suppressed column holds four records. The Bybee and Yoo OLC memos from 2002 to 2009, suppressed under rule-one failure — opinion-shop in a non-record-disciplined system. Boeing's internal MCAS records before the 2021 DPA, suppressed under rule-two failure — corporate-counsel privilege framed the chain as single-party. The second-term DOJ J6 termination memos of January and February 2025, partially surviving under contested rule-three — destruction made visible by prosecutor resignation letters rather than by institutional inventory. The DOGE Signal and records destruction of early 2025, contested under rule three in CREW versus DOGE and American Oversight versus DOGE. One row sits outside the rules' scope: the Russian Decree 330 and 585 registries on Ukrainian children from 2022 onward, where the counter-record from outside — OHCHR HRMMU, the Joint Investigation Team, the European Court of Human Rights, the ICC, and the Children of War Foundation — is the operative mechanism rather than any rule internal to the laundering jurisdiction. End of audio rendering.

::: {.figure-embed}
[〔图：Ten record fights in two columns: every record that survived satisfied at least two of the three rules — write it down, keep custody outside, make destruction visible — and every suppressed record lacked at least two.〕](https://github.com/xiaolai/no-one-did-it/blob/b5e9c8f820bc04bb1d452d4109318673c08c3945/book/chapters-v6/book/evidence/diagrams/proofs/ch12-record-fights.png)
:::

The differences between the surviving cases and the suppressed cases are not random. They are the rules. In every survived row, the record satisfied at least two of the three rules; in every suppressed row, at least two of the rules were absent or partial.

The asymmetry IS our argument. We have documented enough of both columns to support the empirical claim. The rules are not our invention. They are the features of the cases in the surviving column, abstracted into portable form.

The asymmetry table also fixes the rules' scope condition. The three rules operate on institutions acting within a non-adversarial legal order — one in which interceptors (courts, inquiries, watchdog NGOs, journalists, opposition successors) can compel disclosure or assemble counter-records that the originating institution must answer. The first nine rows of the table sit inside that scope. The Russian Decree row does not. In adversary-controlled jurisdictions, rules internal to the adversary do not operate at all; the operative mechanism shifts to counter-record from outside — OHCHR HRMMU, the Joint Investigation Team, the European Court of Human Rights, the International Criminal Court, and externally-domiciled registries such as the Children of War Foundation's. Our three rules are addressed to institutions whose record-keeping behaviour is subject to external interception. They are not a universal claim. The adversary-jurisdiction case is the boundary that names the rules' reach.

Record discipline is the difference between a counter-case and a laundering. We have documented enough of both, inside the scope where the rules can operate, to support the claim.

Write, log, make destruction visible

Stated as a protocol, in three movements.

> When a decision is made, write it down on the day it is made, with the named decider's signature.
>
> When a record moves between custodians, log the move with timestamp and signature. At least one custodian must sit outside the accuser's reporting line.
>
> When a record is to be destroyed, log the destruction order *before* the destruction, with the named authoriser, the inventory of records destroyed, and the legal basis for the destruction.

The three rules together name the record as responsibility laundering's primary battlefield. Hold the battlefield, and chapter 11's design rule of named attribution becomes operative. Lose the battlefield, and every named responsible party in chapter 11 holds named responsibility in name only. The rules do not produce certainty. They produce the conditions under which the cases in our surviving column became visible.

The rules are not new. The Federal Records Act has been in force since 1950, the Presidential Records Act since 1978, the Freedom of Information Act since 1966, the FOIA Improvement Act of 2016, the UK Public Records Acts since 1958 and 1967, the FOI Act 2000, the Inquiries Act 2005, and a body of state archival statutes across many jurisdictions. The deficit is not the statute. The deficit is the enforcement architecture — the routine, real-time, non-litigation-triggered application of destruction-visibility, multi-custodian chain-of-custody, and write-it-down-on-the-day discipline to the records that institutions hold against the populations they regulate.

The shape of the gap is concrete. In *Judicial Watch v. NARA*, the D.C. Circuit confronted directly the question of whether the Federal Records Act's enforcement scheme allows a private suit to compel the Archivist to refer a records-destruction matter to the Attorney General for action under 44 U.S.C. § 3106.[¹⁶](#c12-16) The court's reading of the statute left the enforcement layer dependent on an Archivist's referral the statute makes discretionary, and on an Attorney General sitting in the same executive branch whose records destruction is at issue. The destruction-visibility provisions chapter 12 advocates exist in the statute as nominal duties. They lack the standing, the timetable, and the external trigger that would make routine non-litigation enforcement operate. The Federal Records Act amendments enacted after the 2017 White House records-practice disclosures tightened some provisions and left the enforcement architecture substantially unchanged. We advocate the enforcement architecture itself — the missing layer between statute and remedy — not new statutory text.

Four records that survived

Four cases, each our evidence that the three rules work in practice. The cases are drawn from the surviving column of the asymmetry table. Each panel names the moment of survival, the named record-keeper or interceptor, the year, and the mechanism. Each panel names the cost of escape. We do not soften the cost.

##### The Pentagon Papers (1971).

The moment of survival was the publication decision at *The New York Times* and *The Washington Post* in June 1971 and the Supreme Court's per curiam ruling against prior restraint on 30 June 1971 (403 U.S. 713). The named record-keepers were Daniel Ellsberg, who photocopied the study across late 1969, and Anthony Russo, who helped operate the photocopier. The named interceptors at the publication layer were the editorial leadership of the *Times* and the *Post*. The mechanism was preservation outside the originating institution. Copies existed where the Department of Defense could not recall them. The Department of Justice could attempt prior restraint, but the attempt was forced into a venue — the federal courts — that produced an adversarial record of its own. The cost of escape was personal. Ellsberg and Russo faced charges under the Espionage Act of 1917. The charges against Ellsberg were dismissed on May 11, 1973, by Judge Byrne, on grounds of government misconduct — the Fielding break-in. The transferable lesson is sharp. Preservation outside the originating institution converts a suppressible record into a non-suppressible one. The preservation has personal cost. The cost is part of the lesson.

##### The Iran-Contra reconstruction (1986–1993).

The moment of survival was the November 28, 1986 preservation order issued by career staff at the White House Communications Agency, directing that PROFS backup tapes be retained rather than recycled. The Tower Commission Report establishes the dates of the WHCA preservation order and the destruction acts; the record does not establish whether the two sets of actions were undertaken in awareness of each other, and these two events appear in the record in this order rather than in a documented causal relation. Oliver North's documented shredding of November 1986 and John Poindexter's destruction of contemporaneous notes had already occurred; the backup-tape recycle schedule, had it operated, would have completed the destruction across the following weeks. The named record-keepers were the WHCA career staff who issued the preservation order. The Tower Commission, reporting on February 26, 1987, was the first interceptor to use the preserved tapes, reprinting hundreds of PROFS notes in its report. Independent Counsel Lawrence E. Walsh and his forensic team built on that recovery across 1987–1992, extracting further volumes of contemporaneous communication from the same architecture. The mechanism was a chain-of-custody preservation decision made at a layer the destroyers did not control: the backup infrastructure was operated by an executive-branch communications office whose career staff acted on standard retention judgement before the institutional pressure to recycle could overwhelm the recoverable record. The cost of escape was substantial. The Tower Commission ran four months to publication; the Walsh investigation ran nearly seven years. Walsh's December 24, 1992 statement that "the Iran-contra cover-up, which has continued for more than six years, has now been completed" registered the closing of the criminal-accountability track on the day of the Bush 41 pardons of Weinberger and five others. The transferable lesson is sharp. The operative rule-3 escape here was not Walsh's later forensic work. It was a career-staff preservation decision made days after the destruction order, at a layer the destroyers had not reached. Reconstruction is expensive and incomplete. Executive-layer criminal accountability did not follow.

##### The Iraq Inquiry Cabinet papers (2009–2016).

The moment of survival was the Iraq Inquiry's compelled access to UK Cabinet papers under section 21 of the Inquiries Act 2005, across the inquiry's lifespan from establishment on 15 June 2009 through publication of the *Report of the Iraq Inquiry* on 6 July 2016. The named record-keeper was the Cabinet Office, which held the papers. The named interceptor was Sir John Chilcot, chair of the Inquiry. The mechanism was statutory compulsion. Section 21 of the Inquiries Act 2005 empowers an inquiry to require, by notice, the production of documents from any person; the section's compulsory-evidence power is the design move chapter 11 named at length and the design move chapter 12 ports into the record layer.[¹⁷](#c12-17) The cost of escape was the seven-year inquiry timeline and the well-documented Maxwellisation dispute through which subjects of provisional criticism contested the report's drafting before publication. The publication delay itself became a political dispute. The transferable lesson runs against any reader expecting speed. Statutory compulsion works at the pace of the political-legal system, not at the pace of the underlying events. The rule three move here is the structural variant: make classification and disclosure-refusal visible.

##### The Hofeller files (2018–2019; chapter 6).

The moment of survival has three documented dates, which appear in the record in this order rather than in a documented one-to-one causal relation. On July 12, 2019, the Wake County Superior Court admitted 35 of the late Thomas B. Hofeller's redistricting files into the trial record of *Common Cause v. Lewis*. The bench trial ran July 15–26, 2019. On September 3, 2019, the three-judge panel ruled that the challenged North Carolina state legislative maps were unconstitutional partisan gerrymanders, relying in part on the Hofeller materials alongside the plaintiffs' political-science expert testimony, legislator depositions, and the public legislative record; the panel's opinion identifies multiple evidentiary strands, of which the Hofeller files were one. The named record-keeper, in the design-defying sense, was Hofeller himself, who had retained the files on personal hard drives. The named interceptor was Stephanie Hofeller Lizon, his estranged daughter, who discovered the drives after his death in August 2018 and provided them to the plaintiffs in *Common Cause v. Lewis* through their counsel. The mechanism was a posthumous custody change — an event that no institutional design controlled. The files demonstrated the redistricting architecture that our chapter 6 census-citizenship anchor had only suspected. The cost of escape was the precariousness of the disclosure itself. Had Hofeller destroyed the drives before his death, had his daughter not discovered them, had she not provided them to *Common Cause v. Lewis*, had the litigation not been live at the moment of discovery, the architecture would have remained unknown. The transferable lesson is the honest closer. The most consequential disclosures sometimes turn on events the design cannot reach. The rules increase the probability of survival. They do not guarantee it. We are honest about this. The rules are necessary. They are not sufficient.

Our reader-level test is administration-agnostic, institution-agnostic, and role-specific. Wherever any of us is part of an institutional record — making it, holding it, requesting it, releasing it, or finding it has gone missing — we are inside the argument. The recognition trigger is simple. The record we would need to prove the case is in the hands of the party making the case against us. A destruction announcement is made without a destruction inventory. A record-keeping schedule is invoked without naming the specific records it covers. Two of three is high-confidence. Three of three names a record-control laundering attempt; expect to need an external intercepting mechanism.

The single move that holds across every position rhymes on one principle: put the record somewhere the interested party cannot reach. Replicate the document the originating institution would prefer buried, store the copy beyond that institution's reach, and — where it can be published — name which other custodians hold copies, so the published piece is itself a chain-of-custody artefact. The prior-restraint motion the *Times* faced in *New York Times Co. v. United States* in 1971 is part of the discipline, not a reason to fold. The cheapest version of the move, available to anyone with the time, is to ask for the *index* of records before the records themselves — because where no index can be produced the records were never under discipline — and to keep a dated copy of the request and the response.

The danger to refuse is sole custody. We are set up as the goat the moment we are made the only keeper of a record whose loss would serve another party — so insist on a second custodian before the record matters: an external archive, a personal copy under our own access controls, an institutional archive outside our reporting line. Where the second custodian is refused, the refusal is itself the finding. The whole chapter reduces to one question. Where can more than one party reach the record we would need? If the answer is nowhere, the seam is already named.

Chapter 13 picks up the same question and returns it as a portable instrument.

References

[^425]: Lawrence E. Walsh, Final Report of the Independent Counsel for Iran/Contra Matters, vol. I, ch. 5 (Fawn Hall) — the November 1986 shredding, alteration, and removal of documents from the office of Lt. Col. Oliver L. North. [irp.fas.org](https://irp.fas.org/offdocs/walsh/chap_05.htm), [archived snapshot](http://web.archive.org/web/20260509155645/https://irp.fas.org/offdocs/walsh/).
[^426]: Lawrence E. Walsh, Final Report of the Independent Counsel for Iran/Contra Matters, vol. I, ch. 3 (United States v. John M. Poindexter) and ch. 2 (United States v. Oliver L. North) — North's deletion of 736 PROFS messages and Poindexter's of 5,012, Nov. 22–29, 1986, later recovered from backup tapes. [irp.fas.org](https://irp.fas.org/offdocs/walsh/chap_03.htm), [archived snapshot](http://web.archive.org/web/20260509155645/https://irp.fas.org/offdocs/walsh/).
[^409]: New York Times Co. v. United States, 403 U.S. 713 (1971), 403 U.S. 713, 714 (per curiam opinion). [law.cornell.edu](https://www.law.cornell.edu/supremecourt/text/403/713), [archived snapshot](http://web.archive.org/web/20260510052022/https://www.law.cornell.edu/supremecourt/text/403/713).
[^410]: New York Times Co. v. United States, 403 U.S. 713, 717 (1971) (Black, J., concurring), 403 U.S. 713, 717 (Black, J., concurring). [law.cornell.edu](https://www.law.cornell.edu/supremecourt/text/403/713), [archived snapshot](http://web.archive.org/web/20260510052022/https://www.law.cornell.edu/supremecourt/text/403/713).
[^411]: Memorandum from Jay S. Bybee, Asst. Atty. Gen., OLC, to Alberto R. Gonzales, Counsel to the President, "Standards of Conduct for Interrogation under 18 U.S.C. §§ 2340–2340A" (Aug. 1, 2002); and DOJ Office of Professional Responsibility, "Report: Investigation into the Office of Legal Counsel's Memoranda Concerning Issues Relating to the Central Intelligence Agency's Use of 'Enhanced Interrogation Techniques' on Suspected Terrorists" (July 29, 2009), Bybee memo Aug. 1, 2002; OPR Report July 29, 2009 (final report released Feb. 19, 2010). [irp.fas.org](https://irp.fas.org/agency/doj/opr-final.pdf).
[^412]: Federal Rule of Evidence 901 — Authenticating or Identifying Evidence, Fed. R. Evid. 901(a). [law.cornell.edu](https://www.law.cornell.edu/rules/fre/rule_901), [archived snapshot](http://web.archive.org/web/20260501113732/https://www.law.cornell.edu/rules/fre/rule_901).
[^413]: Federal Rule of Civil Procedure 37(e) — Failure to Preserve Electronically Stored Information (as amended December 1, 2015), Fed. R. Civ. P. 37(e) (as amended Dec. 1, 2015). [law.cornell.edu](https://www.law.cornell.edu/rules/frcp/rule_37), [archived snapshot](http://web.archive.org/web/20260401215246/https://www.law.cornell.edu/rules/frcp/rule_37).
[^414]: Federal Records Act, 44 U.S.C. ch. 31, §§ 3101 et seq. (Records Management by Federal Agencies), 44 U.S.C. ch. 31, §§ 3101–3107 (especially § 3106). [law.cornell.edu](https://www.law.cornell.edu/uscode/text/44/chapter-31), [archived snapshot](http://web.archive.org/web/20260401115708/https://www.law.cornell.edu/uscode/text/44/chapter-31).
[^415]: Citizens for Responsibility and Ethics in Washington v. U.S. DOGE Service, No. 1:25-cv-00511 (D.D.C., Judge Christopher R. Cooper) — complaint filed Feb. 20, 2025; preliminary injunction directing preservation of DOGE records and ruling that USDS is likely subject to FOIA, March 10, 2025, CREW v. U.S. DOGE Service, No. 1:25-cv-00511, Mem. Op. and Order, Mar. 10, 2025 (D.D.C.) (Cooper, J.); Order Denying Motion to Reconsider, Mar. 20, 2025. [courtlistener.com](https://www.courtlistener.com/docket/69658871/citizens-for-responsibility-and-ethics-in-washington-v-us-doge-service/), [archived snapshot](https://web.archive.org/web/20260526113503/https://www.courtlistener.com/docket/69658871/citizens-for-responsibility-and-ethics-in-washington-v-us-doge-service/).
[^416]: Ethics in Government Act of 1978, Pub. L. 95-521, Title VI, codified at 28 U.S.C. §§ 591–599 (Independent Counsel provisions), as in force in December 1986 at the time of Lawrence E. Walsh's appointment as Independent Counsel for Iran/Contra Matters, Pub. L. 95-521, Title VI (92 Stat. 1824, 1867 et seq., Oct. 26, 1978); codified at 28 U.S.C. §§ 591–599 as in force Dec. 1986. [govinfo.gov](https://www.govinfo.gov/content/pkg/STATUTE-92/pdf/STATUTE-92-Pg1824.pdf), [archived snapshot](https://web.archive.org/web/*/govinfo.gov/content/pkg/STATUTE-92/pdf/STATUTE-92-Pg1824.pdf).
[^417]: Report of the President's Special Review Board (Tower Commission Report), Feb. 26, 1987, Tower Commission Report, Feb. 26, 1987; Part III (Arms Transfers to Iran and Contra Diversion) and Appendix materials reprinting PROFS notes. [archive.org](https://archive.org/stream/TowerCommission/President%27s%20Special%20Review%20Board%20%28%22Tower%20Commission%22%29_djvu.txt), [archived snapshot](https://web.archive.org/web/20260527002606/https://archive.org/stream/TowerCommission/President%27s%20Special%20Review%20Board%20%28%22Tower%20Commission%22%29_djvu.txt).
[^418]: White House Communications Agency (WHCA) decision of Nov. 28, 1986 to preserve National Security Council PROFS (Professional Office System) electronic-mail backup tapes rather than recycle them under the standard retention schedule, as documented in the Report of the President's Special Review Board (Tower Commission Report, Feb. 26, 1987) and elaborated in the Final Report of the Independent Counsel for Iran/Contra Matters (Walsh, Aug. 4, 1993, Vols. I–III), Tower Commission Report, Feb. 26, 1987 (Part III appendices on PROFS-note recovery); Walsh Final Report, Aug. 4, 1993 (Vol. I narrative discussion of PROFS-backup-tape recovery). [irp.fas.org](https://irp.fas.org/offdocs/walsh/), [archived snapshot](http://web.archive.org/web/20260509155645/https://irp.fas.org/offdocs/walsh/).; Report of the President's Special Review Board (Tower Commission Report), Feb. 26, 1987, Tower Commission Report, Feb. 26, 1987; Part III (Arms Transfers to Iran and Contra Diversion) and Appendix materials reprinting PROFS notes. [archive.org](https://archive.org/stream/TowerCommission/President%27s%20Special%20Review%20Board%20%28%22Tower%20Commission%22%29_djvu.txt), [archived snapshot](https://web.archive.org/web/20260527002606/https://archive.org/stream/TowerCommission/President%27s%20Special%20Review%20Board%20%28%22Tower%20Commission%22%29_djvu.txt).
[^419]: Report of the President's Special Review Board (Tower Commission Report), Feb. 26, 1987, Tower Commission Report, Feb. 26, 1987; Part III (Arms Transfers to Iran and Contra Diversion) and Appendix materials reprinting PROFS notes. [archive.org](https://archive.org/stream/TowerCommission/President%27s%20Special%20Review%20Board%20%28%22Tower%20Commission%22%29_djvu.txt), [archived snapshot](https://web.archive.org/web/20260527002606/https://archive.org/stream/TowerCommission/President%27s%20Special%20Review%20Board%20%28%22Tower%20Commission%22%29_djvu.txt).
[^420]: Lawrence E. Walsh, Independent Counsel, Final Report of the Independent Counsel for Iran/Contra Matters, Vols. I–III (Aug. 4, 1993); and Walsh contemporaneous statement of Dec. 24, 1992 on President George H. W. Bush's pardons of Caspar Weinberger and five others, as incorporated into the Final Report, Walsh statement, Dec. 24, 1992; reprinted in Final Report of the Independent Counsel for Iran/Contra Matters, Vol. I, Part XI (Concluding Observations) / discussion of the Bush 41 pardons (Aug. 4, 1993). [irp.fas.org](https://irp.fas.org/offdocs/walsh/), [archived snapshot](http://web.archive.org/web/20260509155645/https://irp.fas.org/offdocs/walsh/).
[^421]: Lawrence E. Walsh, Independent Counsel, Final Report of the Independent Counsel for Iran/Contra Matters, Vols. I–III (Aug. 4, 1993); and Walsh contemporaneous statement of Dec. 24, 1992 on President George H. W. Bush's pardons of Caspar Weinberger and five others, as incorporated into the Final Report, Walsh statement, Dec. 24, 1992; reprinted in Final Report of the Independent Counsel for Iran/Contra Matters, Vol. I, Part XI (Concluding Observations) / discussion of the Bush 41 pardons (Aug. 4, 1993). [irp.fas.org](https://irp.fas.org/offdocs/walsh/), [archived snapshot](http://web.archive.org/web/20260509155645/https://irp.fas.org/offdocs/walsh/).
[^422]: Judicial Watch line of Federal Records Act enforcement litigation against the National Archives and Records Administration, including Judicial Watch, Inc. v. National Archives and Records Administration (D.C. Cir.), reading the FRA's enforcement scheme in light of Armstrong v. Bush, 924 F.2d 282 (D.C. Cir. 1991) and Armstrong v. Executive Office of the President, 1 F.3d 1274 (D.C. Cir. 1993); the case law leaves enforcement under 44 U.S.C. § 3106 dependent on discretionary Archivist referral to the Attorney General, Armstrong v. Bush, 924 F.2d 282 (D.C. Cir. 1991); Armstrong v. Executive Office of the President, 1 F.3d 1274 (D.C. Cir. 1993); subsequent Judicial Watch v. NARA panel decisions. [courtlistener.com](https://www.courtlistener.com/docket/4970930/judicial-watch-inc-v-national-archives-and-records-administration/), [archived snapshot](http://web.archive.org/web/20260526083458/https://www.courtlistener.com/docket/4970930/judicial-watch-inc-v-national-archives-and-records-administration/).
[^423]: Inquiries Act 2005 (UK), c. 12, §21. [legislation.gov.uk](https://www.legislation.gov.uk/ukpga/2005/12/section/21), [archived snapshot](http://web.archive.org/web/20260526115214/https://www.legislation.gov.uk/ukpga/2005/12/section/21).

Chapter 13 — A Reader's Field Guide

---
chapter_id: 13-a-readers-field-guide
chapter_number: 13
part: IV
title: A Reader's Field Guide
evidence_grade: A
---

Chapter 13 — A Reader's Field Guide

> Is virtue a thing remote? I wish to be virtuous, and lo! virtue is at hand.
>
> — Confucius, *Analects* 7.29, trans. James Legge

Twelve chapters ago we opened on two goats brought before a priest at the entrance of a tent in the wilderness. One was killed. The other had the iniquities of the community confessed over its head and was sent into the desert with a designated man, never to return. The community watched. The priest spoke. The substitution was named.

What the modern case lost, we argued there, was the naming. The altar moved. The goats kept arriving — a Dutch council communist in a burning Reichstag, an oil-field worker on a Therac-25 treatment couch, two Italian immigrants on a streetcar in Brockton, a flight crew over Bishoftu, a sub-postmaster in a Surrey courtroom, a family separated under a misdemeanor statute, the named cause that was "the data," "the model," "the process," "the procedure." The substitution went underground into the morning headline, the press release, the docket number, the corporate apology, the version-number readme. The community no longer watched a public rite. The community read the official statement and moved on.

We have reached this point after twelve chapters in the position the ancient witness occupied: watching the rite. The next move, in the ancient rite, was the priest's. There is no priest here. There is a field guide.

The instrument is ours.

Three reasons to leave it on the shelf

Three reasons are routinely given for why a diagnostic of this kind should live inside a volume — read once, admired, shelved — rather than walk out into working life. Each has a strong form. We set each at its strongest, then set the artefact next to it.

The first is that a serious analytic instrument requires expertise a general audience does not have. The eight questions are not a quiz; the four categories are not a Buzzfeed test; the responsibility chain is law-school work in disguise. Put crudely: leave it to the analysts. The diagnostic in untrained hands will be mechanical, miscalibrated, and confident in the wrong places.

The strong form of the objection is real and we do not deny it. The second-week law student who reaches for the term "war crime" before the procedural stage supports it produces exactly the kind of mis-attribution the diagnostic was built to interrupt. The diagnostic in untrained hands can produce its own laundering — a public verdict bolted onto a partial reading.

Our answer is structural and is built into the artefacts themselves. The six artefacts that follow are not a checklist. Each carries, on its face, a chapter pointer to the case at which the artefact was installed. When we find a cell unclear, a category contested, or a layer unfamiliar, we return to the worked example. The field guide is the portable reminder; the worked example is the discipline. The two travel together.

The second reason is that the diagnostic is for institutional actors — judges, inspectors general, parliamentary staffers, oversight-committee members — and translates poorly to the citizen reading the morning headline. The strong form: the eight questions name the analytic task of a public inquiry. The citizen is not running a public inquiry.

Our answer is that the citizen, the juror, the journalist, the employee, the manager, the contractor, the professional, and the voter each meet institutional events at a specific surface. The crosswalk — Artefact 5 — names that surface. The eight questions do not change at each surface; the action does. What the citizen does at the seam of an algorithmic denial is not what the voter does at the seam of an inquiry-statute amendment, and the crosswalk names the difference rather than collapsing it.

The third reason is that the diagnostic is its argument — that lifting the diagnostic out of its installing chapters strips it of the narrative architecture that made it intelligible. The strong form: anyone who reads only the field guide will arrive at the four categories without the cases that taught the categories, and will reach for the labels without the discipline of the chain. The diagnostic in checklist form becomes a slogan; the slogan becomes a substitute for the work.

This is our load-bearing counterargument, and it returns at full strength below. The structural answer is the same as before, scaled up: the artefacts do not travel alone. A by-chapter index — Artefact 6 — is the tether. Every artefact points back to its installing chapter. When we find the artefact insufficient for a specific case, we return to the worked example. The field guide is not a substitute for the installing chapters. The field guide is the diagnostic at the moment it leaves our hands and enters our life.

Six artefacts, to be carried

The heaviest slot, and the deliverable. Six artefacts follow, each designed to be photocopied, folded, and carried.

Artefact 1 — The eight questions

The diagnostic is verbatim from our project file. Read them in order. Write down the answers. Return to the question whose answer is "unknown."

> For every case ask:
>
> 1. Who or what was publicly blamed?
> 2. Who had control?
> 3. Who benefited?
> 4. Who knew or should have known?
> 5. Who could have prevented recurrence?
> 6. Who controlled the record?
> 7. Who bore the cost?
> 8. What would responsibility look like if it followed control instead of visibility?

How to use them. Take a single institutional event — a court filing, a news report, a corporate apology, an inquiry charter, an AI incident report, a contract clause. Walk the eight questions in order. Where the public account names a non-human verb-subject — the model, the system, the process, the institution — the answer to question 1 is the alibi container, and the answer to question 8 names the human chain the container displaced. The mismatch between the answers to question 1 and question 8 is the laundering's location. The mismatch is what the next five artefacts unpack.

The eight questions were installed in chapter 3 against the Volkswagen Dieselgate case and the Ford Pinto case. Both were partial scapegoats in which engineers were named while the executive chain that approved the design choice remained outside criminal exposure. We return to that worked example in chapter 3 when the questions do not yield a clean climb.

Artefact 2 — The four-category taxonomy

The four categories are verbatim from rule `01-case-taxonomy.md` of our project file. They are mutually exclusive; the only allowed mixing is explicit declaration.

> **Pure scapegoat:** mostly innocent actor is blamed to close the case.
>
> **Partial scapegoat:** actor is involved or guilty at one level, but blame stops too low and protects a wider chain.
>
> **System/object alibi:** A tool, machine, algorithm, benchmark, corporation, legal entity, committee, market, or procedure becomes the named cause.
>
> **Cost-bearing goat:** victim or worker is not accused as guilty but absorbs harm while responsibility is avoided.

How to use them. After the eight-question walk, ask which category the laundering fits. The most diagnostic question per category is different. For pure scapegoat, the diagnostic is question 4 (who knew, and what did the documentary record show they knew about the named actor's non-involvement). For partial scapegoat, it is question 2 (who had control above the named actor). For system/object alibi, it is question 1 itself (the named cause is a thing, not a person). For cost-bearing goat, it is question 7 (the cost-bearer is named in the public account; the controller is not).

Hybrid cases exist. Chapter 9's AI training-data litigation cluster is named as system/object alibi (primary) plus cost-bearing goat (secondary): the laundering grammar renames human creative work as substance, and the rights-holders absorbing the cost are not visible at the narrative centre. The hybrid is declared rather than fudged. The taxonomy survives because every claim of "complicated" routes through an explicit second category.

The four categories were installed in chapter 2 against Therac-25 and Boeing 737 MAX (system/object alibi), Sacco and Vanzetti (pure scapegoat), Abu Ghraib military police (partial scapegoat), and Ukrainian children transferred from occupied territory (cost-bearing goat). We return to those worked examples in chapter 2 when the category is contested.

Artefact 3 — The AI-stack three-layer diagnostic

Chapter 9 installed the AI-stack model alibi at three structurally distinct layers. The artefact lays them out in a single column.

> **Input layer.** The named verb-subject is "data," "the corpus," "the shadow library," "the market." The displaced human decisions are the named acquisition decisions: which corpora, on what dates, from what sources, under what licensing documentation. The diagnostic question is the acquisition record. The chapter-9 worked example is the *Bartz v. Anthropic* summary judgment of 23 June 2025, in which Judge Alsup distinguished lawfully acquired books from shadow-library copies on the basis of the acquisition record.
>
> **Deployment layer.** The named verb-subject is "the model," "the reward signal," "the release process." The displaced human decisions are the version designation, the post-training reward-shaping log, and the named individuals who approved the default-model swap. The diagnostic question is the post-training change record. The chapter-9 worked example is OpenAI's GPT-4o sycophancy post-mortem of 29 April 2025, which named the optimisation framework rather than any individual.
>
> **Evaluation layer.** The named verb-subject is "the variant," "the policy," "the benchmark." The displaced human decisions are the submission decision and the disclosure-policy interpretation. The diagnostic question is the benchmarked-versus-shipped variant gap. The chapter-9 worked example is the Llama-4-Maverick-03-26-Experimental submission to LMArena in April 2025 and LMArena's subsequent policy revision.

How to use them. Run the eight questions three times per AI system — once at each layer. The named human decision-maker is different at each layer. The displaced decision is different. The interception instrument is different. Subtract any layer and the alibi sustains at the unexamined floor.

Artefact 4 — The five-layer war stack

Chapter 10 installed the war-laundering stack at five layers. The artefact lays them out in a single column.

> **Chain-of-command secrecy layer.** A wartime incident is publicly blamed on enlisted personnel or a low-rank unit while doctrine memos, theatre-commander authorisations, and interrogation-technique approvals remain classified during the operative window. The chapter-10 worked example is the Abu Ghraib chain: night-shift military police were court-martialled while the OLC memoranda of 1 August 2002 and the Sanchez authorisation of 14 September 2003 were declassified only after the operative window had closed.
>
> **Proxy-deniability layer.** Operations are conducted by entities with deniable legal status — armed groups, people's republics, local forces, irregulars — while sovereign-state command of supply, training, and recovery is publicly denied. The chapter-10 worked example is MH17: the Buk-TELAR belonged to the Russian 53rd Anti-Aircraft Missile Brigade; the public framing attributed the shootdown to local armed groups or to Ukrainian forces.
>
> **Contractor legal-status-shell layer.** Private military, private surveillance, private interrogation, and private logistics operate under contractual arrangements that exclude criminal jurisdiction and are described as administrative convenience rather than as accountability architecture. The chapter-10 worked example is Blackwater / Nisour Square: Coalition Provisional Authority Order 17 excluded Iraqi criminal jurisdiction; the State Department contract insulated the corporate principal.
>
> **Civilian-invisibility-through-euphemism layer.** Cost-bearers — civilians, refugees, transferred children, displaced populations — are described in care-language ("evacuation," "guardianship," "humanitarian relocation," "rescue from a war zone") while the underlying conduct continues. The chapter-10 worked example is the Russian transfer of Ukrainian children under Presidential Decrees No. 330 and No. 585 of 2022.
>
> **Strategic-justification layer.** A national institutional product — a national intelligence estimate, an expert assessment, a model output — is presented in public with hedges stripped, and the body that audits the product is barred by charter from examining its policy use. The chapter-10 worked example is the Iraq WMD National Intelligence Estimate of October 2002, presented as certainty at the UN Security Council, and the Robb-Silberman Commission's Executive Order 13328 charter that excluded examination of policy use.

How to use them. Ask which layers are firing simultaneously. Where three or more of the five are present, the case is structurally laundering-shaped, and the diagnostic applies. Treat any claim that responsibility in war is irrecoverable as the laundering mechanism itself.

Artefact 5 — The role-by-domain crosswalk

The crosswalk is our largest artefact. Rows are roles. Columns are six anti-laundering moves we have installed across chapters 2 through 12. Each cell names one specific move the named role can execute in the named domain. Where the same move applies across multiple cells, the cell points across to the shared move rather than restating it.

The six anti-laundering moves (columns), in installation order:

> **Move 1 — Name the controller.** When a procedure or institution is cited as cause, ask in writing which named person made the decision the procedure produced. (Chapters 2, 3, 5, 6.)
>
> **Move 2 — Demand the acquisition or decision record.** Ask for the document — the manifest, the email chain, the approvals log, the Statement of Responsibilities — that shows who decided what, on what date, under what alternative available to them. (Chapters 7, 10, 12.)
>
> **Move 3 — Refuse the signature without the chain.** Refuse the named-signer position for a function whose written allocation, sub-delegations, and escalation paths have not been documented. (Chapters 6, 7, 11.)
>
> **Move 4 — Make the destruction or classification visible.** Where records are destroyed, classified, sealed, or deleted, demand the destruction log, the classification basis, or the retention exception in writing. (Chapters 9, 10, 12.)
>
> **Move 5 — Apply the design test before signing the design.** When proposing, signing, or advocating an institutional design, ask whether a named individual bears the consequence of the control question failing at each layer the design touches. (Chapter 11.)
>
> **Move 6 — Vote and advocate for the compelling architecture.** Where the design choice is structural — inquiry statutes, oversight charters, AI disclosure regimes, certification regimes, FOIA reforms — choose the architecture that grants compulsory-evidence access over the architecture that fences it. (Chapters 8, 11, 12.)

The crosswalk:

::: {.figure-embed}
[〔图：Eight reader roles against six anti-laundering moves: the eight questions do not change at each reader's surface — the action does — and the blank cells are deliberate.〕](https://github.com/xiaolai/no-one-did-it/blob/b5e9c8f820bc04bb1d452d4109318673c08c3945/book/chapters-v6/book/evidence/diagrams/proofs/ch13-crosswalk.png)
:::

> **Citizen.** Move 1: read public-cause language for the verb-subject; name the human chain it displaced (anchor: ch-2 Boeing 737 MAX "Technical Pilots" / "MCAS" framing; ch-6 family separation "the statute required it"). Move 2: file FOIA, subject-access, or public-record requests on the decision the public account left unnamed (anchor: ch-7 *DOJ v New York*, June 27, 2019). Move 3: refuse to accept terms of service, settlements, or institutional consents whose "procedure," "policy," or "system" clauses have no named decider; insist on the named-attribution layer before assent (anchor: ch-11 *Park* doctrine of the responsible corporate officer). Move 4: read for destruction announcements without inventories (anchor: ch-12 PROFS-tape preservation, Nov 28, 1986). Move 5: in public-comment periods, in zoning hearings, and on local-government charters, ask whether the design names the controller or names a procedure (anchor: ch-12 *Common Cause v Lewis* admission of Hofeller files into trial record, Jul 12, 2019). Move 6: vote on inquiry-statute and oversight-charter ballot questions when they appear (anchor: ch-8 Inquiries Act 2005 §21 architecture).
>
> **Juror.** Move 1: in any case where institutional conduct is at issue, ask whether the record presented to the jury was held by a single party to the dispute. Move 2: request, through the court, the production of the underlying acquisition or decision record where the institutional witness's testimony relies on it. Move 3: see Professional row (jurors who are professionals carry their professional duty into deliberation). Move 4: ask whether deleted or unproduced records were preserved under court order. Move 5: not applicable in deliberation. Move 6: not applicable in deliberation.
>
> **Journalist.** Move 1: in any institutional-cause story, name the human chain the public account displaces; refuse the institution's verb-subject (anchor: ch-8 Knight-Ridder Washington Bureau dissent on Iraq WMD, 2002–2003). Move 2: file FOIA / FOI / subject-access requests; pursue the acquisition record, the destruction log, and the procurement document (anchor: ch-10 *Bartz v Anthropic* acquisition-record discovery, leading to the SJ of Jun 23, 2025). Move 3: refuse to publish institutional verb-subjects as authoritative when no named source signs for the decision; require an on-the-record human (anchor: ch-9 administration framings of "the system" vs the dated official statement). Move 4: report the destruction announcement without an inventory as itself the story (anchor: ch-12 OLC memo "withdrawal," Dec 30, 2004). Move 5: cover charter design, not only the inquiry findings (anchor: ch-11 EO 13328 vs Inquiries Act 2005 §21 design asymmetry). Move 6: cover inquiry-statute reform, not only the inquiry it produced (anchor: ch-8 Robb-Silberman scope-fence vs Chilcot scope, contrasted).
>
> **Employee.** Move 1: when cited "policy," "procedure," or "the system" produces an outcome affecting you, ask in writing which named manager made the decision the procedure produced (anchor: ch-5 Boeing Technical Pilots' inferred-knowledge layer). Move 2: keep your own contemporaneous written record of decisions, instructions, and approvals affecting you; do not rely on the institution's record alone (anchor: ch-7 source-ledger discipline; ch-12 chain-of-custody rule). Move 3: refuse to be the named signer for a function whose Statement of Responsibilities you have not been shown (anchor: ch-11 SMCR allocation under FCA SYSC 4.7). Move 4: where institutional records affecting you are scheduled for deletion, request preservation in writing (anchor: ch-12 FRCP 37(e) anticipation-of-litigation duty). Move 5: when your employer asks you to operate within a process, test whether the process names approvers and signatories (anchor: ch-11 SOX §§ 302/906 individual-attestation design). Move 6: where applicable to your jurisdiction, vote and submit public comment on inquiry-statute and oversight-charter questions (anchor: same as Citizen Move 6, ch-8 Inquiries Act 2005 §21).
>
> **Manager.** Move 1: in any institutional decision you sign for, name the controller above you whose authorisation makes the decision yours to sign (anchor: ch-5 chain-of-command analysis at Boeing). Move 2: maintain a written approvals log for decisions you sign; require sub-delegations in writing (anchor: ch-11 SMCR Statement of Responsibilities). Move 3: refuse to sign for an outcome whose Statement of Responsibilities you have not seen (anchor: ch-11 SOX § 302 individual-certification regime). Move 4: where your institution's record-retention policy schedules destruction of records relevant to a known dispute, escalate the spoliation risk in writing (anchor: ch-12 FRCP 37(e); *NYT v OpenAI* preservation order, Nov 2025). Move 5: when you design or amend a process, charter, or contract under your authority, apply the named-attribution test to your own design (anchor: ch-11 Inquiries Act 2005 §21 vs EO 13328 design asymmetry). Move 6: where applicable, vote and submit public comment on inquiry-statute reform; otherwise n/a in this role.
>
> **Contractor.** Move 1: in contract negotiation, name the controller at the principal whose decision triggers your contractual obligation. Move 2: insist that consequential records — performance, payment, change-orders, audit — be held by more than one custodian, including your own. Move 3: refuse "designated person" clauses that leave the name as a placeholder; refuse indemnification clauses that route to the legal entity without naming the officer whose decision triggers the indemnity. Move 4: where the contract authorises destruction of records relevant to a known dispute, escalate. Move 5: read the contract for Park-shaped and SOX-shaped clauses; propose the missing named-attribution clauses in writing. Move 6: not applicable.
>
> **Professional** (licensed practitioner: lawyer, doctor, engineer, accountant, certifying officer, regulated-sector senior manager). Move 1: in any certification, opinion, or attestation, name the sub-delegation document that authorises the decision you are asked to certify. Move 2: maintain your own contemporaneous record of the institutional decisions on which your professional judgment relies. Move 3: refuse to certify, opine, or attest on the strength of an institutional product whose acquisition, deployment, or audit record you cannot inspect. Move 4: where retention policies threaten destruction of records relevant to your professional duty, escalate in writing. Move 5: read the regulatory regime under which you certify for design absences; flag them in writing to your regulator. Move 6: vote and submit comment on professional-regulation rule-makings.
>
> **Voter.** Move 1: in election coverage and public debate, read political accounts of institutional events for the displaced verb-subject; refuse the institution's framing (anchor: ch-9 administration statements characterising removed individuals as a category rather than as individuated subjects of process). Move 2: ask whether the inquiry whose findings are cited in the political account had compulsory-evidence access (anchor: ch-8 Chilcot Inquiries Act §21 access to Cabinet papers vs Robb-Silberman fenced scope). Move 3: refuse to sign ballot initiatives, constitutional amendments, or candidate endorsements whose stated mechanism names a procedure but not the responsible officer (anchor: ch-11 Park doctrine; named-attribution as voter test). Move 4: ask whether the agency or institution whose conduct is at issue has destroyed records under retention policy during the operative window (anchor: ch-9 CREW v US DOGE Service preservation order, Mar 10, 2025). Move 5: read draft inquiry statutes, FOIA-reform bills, AI disclosure regulations, and oversight charters for named-attribution specifications (anchor: ch-12 *Armstrong v Bush* FRA-enforcement architecture). Move 6: vote and submit comment for the compelling architecture — the inquiry chair with compulsory-evidence power, the certifying officer with personal exposure, the FOIA regime with no agency-status carve-out, the AI training-data disclosure obligation with no trade-secret blanket exception (anchor: ch-11 four design moves).

The bottom-right corner cell — Voter, Move 6 — is the most consequential, because the architecture of every other cell is determined by what the voter does at this cell. The Inquiries Act 2005 §21 model is one available shape; the Robb-Silberman Executive Order 13328 model is another. We do not need a verdict on which shape is better in every jurisdiction. We need to know which shape is on the ballot.

Artefact 6 — A by-chapter index

Each chapter, named with its installed mechanism and its installed counter-cases.

> **Chapter 1 — The Altar Moves.** Mechanism: civilisations do not stop sacrificing substitutes; the altar moves and the substitute changes shape. The chapter installs the recognition, not a counter-case.
>
> **Chapter 2 — The Four Goats.** Mechanism: four shapes of laundering — pure scapegoat, partial scapegoat, system/object alibi, cost-bearing goat. Anchor cases: Therac-25; Boeing 737 MAX; Sacco and Vanzetti; Abu Ghraib military police; Ukrainian children transfers. Counter-cases — *Cour de Cassation* quash-without-remand 1906 (Dreyfus, against pure scapegoat); Knapp Commission with subpoena power under Mayor John V. Lindsay's Executive Order No. 11 of 21 May 1970 and Frank Serpico's documentation 1966–1970 (against the bad-apples partial scapegoat); *Bates v Post Office* High Court Group Litigation Order with Mr Justice Fraser's *Horizon Issues* judgment of 16 December 2019 and independent technical examiner Jason Coyne (against system/object alibi); *Bringing Them Home* report of 26 May 1997 and the National Apology of 13 February 2008 (against the cost-bearing-goat euphemism — the Australian Stolen Generations).
>
> **Chapter 3 — Who Could Have Stopped It?** Mechanism: the eight-question diagnostic. Anchor cases: Volkswagen Dieselgate; Ford Pinto. Counter-cases — VW asymmetric interception with the United States v. Liang and United States v. Schmidt named-individual criminal convictions; *Grimshaw v. Ford Motor Co.* civil verdict and NHTSA recall 78V-138.
>
> **Chapter 4 — The Proxy and the Sponsor.** Mechanism: proxy deniability plus legal-person laundering. Anchor cases: MH17 (state-via-proxy); Crimea (state-direct); Blackwater / Nisour Square (private contractor). Counter-cases — Joint Investigation Team Buk-TELAR traceback to the Russian 53rd Anti-Aircraft Missile Brigade and the Hague District Court at Schiphol's 17 November 2022 convictions in absentia of Girkin, Dubinsky, and Kharchenko; ICAO Council finding of 12 May 2025 against the Russian Federation under Article 3 bis of the Chicago Convention; *United States v. Slatten* and *United States v. Slough et al.* DOJ prosecutions (the 22 December 2020 pardons later reversed the individual convictions; the interception is documented as having reached and been reversed).
>
> **Chapter 5 — The Guilty Goat.** Mechanism: a directly culpable low-level actor is named to prevent inquiry from climbing to the parent corporation, the design layer, or the regulator. Anchor cases: Bhopal / Union Carbide; Boeing 737 MAX pilot-blame attempt. Counter-cases — the partial reversed-corporate-veil reading of Bhopal at the Indian Supreme Court (qualified by the un-honoured US extradition of Warren Anderson); KNKT Indonesian and ECAA Ethiopian accident-investigation final reports under ICAO Annex 13, which reached the design layer the pilot-blame framing was designed to halt at.
>
> **Chapter 6 — The Pretext.** Mechanism: a clean legal category is recruited to carry a different policy motive. Anchor cases: family separation under 8 U.S.C. § 1325; the 2020 census citizenship question; the 2019 OMB hold on Ukraine security assistance. Counter-cases — *Ms. L. v. ICE*, Judge Sabraw's federal class action; SCOTUS APA review in *Department of Commerce v. New York*, with the posthumous disclosure of the Hofeller files in North Carolina state proceedings; GAO opinion on the Impoundment Control Act and the first-impeachment record.
>
> **Chapter 7 — The Record Is the Battlefield.** Mechanism: the actor that controls the record controls future responsibility. Anchor cases: the UK Horizon / Post Office Ltd / Fujitsu prosecutions. Counter-cases — Mr Justice Fraser's *Horizon Issues* judgment of 16 December 2019 paired with the Group Litigation Order and Jason Coyne's reverse-engineering of the Known Error Log; the *Hamilton & Others v Post Office Ltd* convictions quashed at the Court of Appeal on 23 April 2021. Callbacks: Therac-25 artefact-layer interception via Leveson and Turner's 1993 reconstruction; MH17 state-layer interception via the JIT.
>
> **Chapter 8 — When Power Calls Itself the Goat.** Mechanism: accountability mechanisms are reframed as themselves the laundering being corrected; dismantling the mechanism is presented as the reform. Anchor cases: the Alien Enemies Act invocation of March 2025 with the Abrego Garcia removal; the January 6 pardons of January 2025; the DOGE / Presidential Records Act litigation. Counter-cases — *Trump v. J.G.G.* per curiam of 7 April 2025 (habeas before AEA removal); *Noem v. Abrego Garcia* per curiam of 10 April 2025 (Article III interception requiring the government to "facilitate" release); Judge Cooper's preliminary injunction and preservation orders in *CREW v. U.S. DOGE Service* in March 2025. Older echoes: the OLC memorandum of 5 August 1974 on the limits of self-pardon (Watergate); the Walsh Final Report of 4 August 1993 on Iran-Contra.
>
> **Chapter 9 — The Model Did It.** Mechanism: the AI-stack model alibi at three layers — input, deployment, evaluation. Anchor cases: the AI training-data litigation cluster; GPT-4o sycophancy; Llama 4 / LMArena. Counter-cases — *Bartz v. Anthropic*, Judge Alsup's summary judgment of 23 June 2025 and the $1.5 billion settlement announced 5 September 2025; *Authors Guild v. OpenAI* discovery proceedings; *NYT v. Microsoft / OpenAI* preservation order of November 2025; *Getty Images v. Stability AI* in the UK High Court, Mrs Justice Smith's territoriality ruling of 4 November 2025; OpenAI's GPT-4o post-mortem of 29 April 2025; LMArena's policy revision of mid-April 2025.
>
> **Chapter 10 — War Is the Perfect Laundry.** Mechanism: war stacks five laundering layers — chain-of-command secrecy, proxy deniability, contractor legal-status shell, civilian-invisibility euphemism, strategic-justification reification. Anchor case: the Iraq WMD intelligence and the public statements that exceeded it. Counter-cases — the Chilcot Iraq Inquiry's seven-year statutory operation under the Inquiries Act 2005 §21 compulsory-evidence power and its 6 July 2016 report; the Hague District Court at Schiphol's MH17 judgment of 17 November 2022; the European Court of Human Rights's *Ukraine and Netherlands v Russia* inter-state proceedings, pending on the merits as of mid-2026.
>
> **Chapter 11 — Make Responsibility Follow Control.** Mechanism: design rule — responsibility follows control by statute, doctrine, or charter construction. Anchor cases: the Park doctrine under the FDCA; Sarbanes-Oxley §§ 302 and 906; the UK Senior Managers and Certification Regime; the Inquiries Act 2005 §21 contrasted with Executive Order 13328. Counter-cases — each design move's documented operation in named cases (*Park / Acme Markets*; SOX certifications across two decades of issuer filings; FCA Final Notices under SMCR; Chilcot's compelled access to Cabinet papers under §21 as the design contrast against the Robb-Silberman fenced charter).
>
> **Chapter 12 — Keep the Record.** Mechanism: three record-discipline rules — write it down; timestamp and chain-of-custody; make destruction visible. Anchor: the three rules walked against chapter 7's five-role conflation, the OLC opinion-shop pattern of chapter 8, and the spoliation jurisprudence around chapter 9's preservation orders. Counter-cases — the Pentagon Papers, Daniel Ellsberg and Anthony Russo 1971, and *New York Times Co. v. United States*, 403 U.S. 713; the Iran-Contra Walsh team's forensic reconstruction of destroyed Poindexter notes from the National Security Council's PROFS backup tapes; the Iraq Inquiry's Cabinet-papers compulsion under Inquiries Act §21; the Pennsylvania Hofeller files surfacing in *Common Cause v Lewis* through posthumous custody change.

The twelve chapters were the explanation. These six artefacts are the instrument. The instrument is now ours.

The chord behind the instrument

The older echo here is the work behind us. The twelve installments this field guide returns to are the chord on which the field guide rests, and we name the chord here so anyone who picks up the field guide first can locate the worked examples.

Chapter 1 named the rite. Chapter 2 named the four shapes. Chapter 3 named the eight questions. Part II — chapters 4, 5, 6, 7 — stress-tested the diagnostic against the proxy, the partial scapegoat at the corporate parent, the administrative-law pretext, and the institution that controls the record. Part III — chapters 8, 9, 10 — stress-tested the diagnostic against the three domains that fight it hardest: war, partisan inversion, and the AI stack. Part IV — chapters 11 and 12 — turned the diagnostic forward into a design rule and a record-discipline triplet.

The field guide is not a thirteenth chapter that adds new mechanism. It is the consolidation pass. Anyone who has not read the twelve chapters reads the field guide and decides which chapter to open first. Anyone who has read them carries the field guide as a portable reminder.

The chord belongs to the twelve chapters, not to the consolidation. The field guide is the chord's printed form.

Audited at the source

Every artefact here has been walked, verified, audited, and gate-passed in a prior chapter. The eight questions are byte-exact to the project's source-of-truth file. The four-category taxonomy is byte-exact to the project's case-taxonomy rule. The AI-stack three layers are taken from chapter 9's anchor argument. The five-layer war stack is taken from chapter 10's anchor argument. The crosswalk cells are built from the beat-10 anti-laundering moves installed in chapters 2 through 12. The by-chapter index draws on each chapter's source ledger.

The discipline that produced the anchors produced the field guide. We have graded every claim A, B, C, or D before it earned a place in the manuscript; the field guide consolidates only A-graded and B-graded material from the originating chapters. The field guide introduces no new citation. Where the artefact summarises a case, the originating chapter's case file and source ledger are the authoritative record.

The instrument's trustworthiness is the discipline of the twelve installments it consolidates.

The reversal: the field guide is not a flattening of the diagnostic. The field guide is the diagnostic at the moment it leaves our hands.

The risk of the checklist

The strongest objection to a field guide of this shape comes from anyone who has finished the twelve chapters and is now being asked to carry six artefacts out into Monday morning.

The objection runs as follows. Field guides flatten the diagnostic into bullet points and lose the structural argument that gave the diagnostic its force. The eight questions on one artefact separated from their cases become a checklist; the four categories on one artefact separated from their installing chapter become labels; anyone who reads only the field guide will apply the diagnostic mechanically and reach wrong answers. Chapter 3 itself flagged this risk inside its own walk — the diagnostic-as-checklist risk. The risk is real at chapter scale. Across twelve chapters it is greater. A field guide that strips the cases from the artefacts converts a diagnostic into a slogan, and the slogan becomes the substitute for the work the twelve chapters asked us to do.

The strong form of the objection has historical evidence on its side. Diagnostic taxonomies that travel without their cases have been weaponised in every direction. "Scapegoat" is now a one-word counter-accusation that does no analytic work and licenses the user to dismiss any contested attribution. "Witch hunt" is its mirror form. The DSM categories travel without their case histories and produce mechanical clinical judgements. Logical fallacies as memes travel without their reasoning context and produce a generation of internet debate that names the fallacy and stops. Each is a diagnostic instrument that became a checklist and then a label. We are right to worry that the field guide on the next photocopy could become the next instance.

We do not deny the risk. We answer structurally. The six artefacts above are not a checklist; each is tethered, by typography and by explicit pointer, to the worked example that installed it. The by-chapter index is the tether. When we find an artefact insufficient for a specific case, we return through the index to the worked example. The artefact is a portable reminder; the worked example is the discipline. The two travel together. Whoever carries only the artefacts will produce shallower diagnostic work than whoever carries both. Whoever carries neither will produce no diagnostic work at all. The field guide raises the floor without raising the ceiling.

The second answer is harder. The diagnostic was never the property of the analyst. We have been arguing for twelve chapters that the laundering depends on the asymmetry between the institutional actor with control and us without it; the diagnostic was assembled, in part, to redistribute that asymmetry. A diagnostic that stays inside the twelve chapters is a diagnostic that has accepted the asymmetry. The risk that the diagnostic in untrained hands will be applied mechanically is real and is smaller than the risk that the diagnostic in trained hands will be applied to the cases the trained hands already control. The field guide accepts the first risk in order to lower the second.

The third answer is the exit move, which we take up below.

The letter that names the model

For twelve chapters we have occupied the seat of the observer. The earlier chapters walked documented intercepts of laundering attempts: the *Cour de Cassation*'s 1906 quash-without-remand of the Dreyfus conviction; the Knapp Commission's subpoena-powered hearings on the NYPD pad system; Mr Justice Fraser's *Horizon Issues* judgment; the Joint Investigation Team's serial-equivalent identification of the Buk-TELAR; Sabraw J's class action ordering reunification of separated families; the SCOTUS APA review of the census citizenship question; Chilcot's compelled access to UK Cabinet papers; the SCOTUS per curiams of April 2025 in *Trump v. J.G.G.* and *Noem v. Abrego Garcia*; the Pentagon Papers publication; Walsh's reconstruction of the destroyed Poindexter notes; Judge Alsup's distinction between lawfully acquired and shadow-library training corpora. Each intercept required an instrument the earlier chapters named. None of the intercepts arrived by itself.

In every named intercept, the intercepting actor was at first a person who noticed something. Picquart noticed that the bordereau handwriting matched Esterhazy's, not Dreyfus's. Serpico noticed that the precinct pad ran on the assumption that he would take. Alan Bates noticed that no other sub-postmaster could possibly have been doing what the Post Office's prosecutors said only he was doing. Hans Blix noticed that UNMOVIC's site visits were finding nothing. The intercepts started before the inquiries existed.

The diagnostic we now hold is the formalisation of what those persons noticed. Our next intercept will not appear in a future edition. It will appear in our working life, in whatever role we hold.

What follows walks one specific shape of next intercept, in the second person, because the shape is most likely to find us in the working life of 2026 and the years after. The shape is the algorithmic denial.

Imagine you receive a letter. The letter cites a model — your bank's credit-risk model; your employer's screening tool; your insurer's pricing engine; a platform's content-moderation classifier; a government agency's eligibility filter. The letter does not name the model's version. The letter does not name the training corpora. The letter does not name the post-training reward-shaping changes that occurred since the last release. The letter names you and an outcome adverse to you. The verb of the letter is *the model decided*. The subject of the verb is the thing.

The eight questions, in order. Who was publicly blamed? You — the addressee of the letter, the bearer of the denial. Who had control? The institution that deployed the model. Who benefited? The institution that deployed the model, in the form of an automated decision at scale. Who knew or should have known? The named human decision-makers at the deployer — the product manager, the risk-committee chair, the model-deployment approver. Who could have prevented? The same. Who controlled the record? The deployer, at all three of chapter 9's layers — the training-corpus manifest at the input layer; the version designation and reward-shaping log at the deployment layer; the procurement-document benchmark gap at the evaluation layer. Who bore the cost? You. What would responsibility look like if it followed control instead of visibility? The chain from the field office that sent the letter, through the operations director who runs the model in production, to the executive who approved the deployment.

You write back. You ask the three-record demand from chapter 9: input-layer (which corpora, on what dates, under what licensing), deployment-layer (which model version, what post-training changes, what appeal path that does not route back to the same model), evaluation-layer (whether the deployed model matches the model whose benchmark scores were cited in any marketing or procurement document supporting the deployment).

The institution may answer. The institution may not. The crosswalk tells you what to do next according to the role you hold. As cost-bearer, you document the denial and the response; the documentation is the record on which a future intercept can rest. As employee — if the model is your employer's — you raise the Statement of Responsibilities question through whatever channel does not route back to the same model. As contractor — if you contract with the deployer — you propose, in writing, the named-attribution clauses the contract did not contain. As professional, you refuse the certification, opinion, or attestation that depends on the institution's product whose record you cannot inspect. As citizen and voter, you submit comment on the AI-disclosure rulemaking the European AI Act and California's SB-942 have begun and the analogous regimes in your jurisdiction have not yet.

The walk above is a shape, not a case. It is hypothetical here by design. The discipline forbids fabricating the specifics of an event that did not happen to a real person. The shape, however, is documented in every direction — the *Bartz* acquisition record at the input layer; the GPT-4o sycophancy post-mortem at the deployment layer; the Llama-4-Maverick-03-26-Experimental submission at the evaluation layer; chapter 7's Horizon five-role conflation that the AI stack reproduces; chapter 11's design rule that the deployer's design did not satisfy; chapter 12's record-discipline triplet the deployer's record did not survive. Whoever reaches the moment the letter arrives meets the earlier chapters at the same time. We carry the field guide into the moment.

The alibi-collapse beat is the moment we recognise the diagnostic operating in our own domain. The diagnostic does not collapse. The alibi the institution sent in the letter collapses. The collapse is partial; the collapse is costly; the collapse is documented. The collapse is real.

Walk the diagnostic before the verdict

The portable rule, stated as protocol:

> Carry the eight questions. Carry the four categories. Carry the three AI-stack layers and the five war-stack layers. Carry the role-by-domain crosswalk. Carry the by-chapter index. When you meet an institutional event that calls for a verdict, walk the diagnostic before forming the verdict. When the diagnostic exposes a gap between the public chain and the responsibility chain, the gap is the chapter already written for you to read.

The rule is portable across the cases installed and the cases not. It applies whether the event is a court filing, a corporate apology, an inquiry charter, an AI incident report, a contract clause, a news framing of a recent institutional event, or an HR investigation in which we are the named party. The rule does not pre-determine the verdict. The rule produces the walk on which the verdict will rest.

This is our one use of the term that names the diagnostic. Responsibility laundering is the recurring institutional method by which power keeps control or benefit while moving blame, liability, moral cost, or public suspicion onto a weaker bearer, procedural shell, proxy, machine, market, committee, legal entity, or record. The diagnostic we now hold is the portable counter-method.

We are the escape

We are the escape.

This is the exit move. For twelve chapters we have been the observer of intercepts. The thirteenth pass inverts the position. The intercepts were performed by people. The intercepts were named — Picquart, Serpico, Alan Bates, Sabraw J, Mr Justice Fraser, the Joint Investigation Team, Chilcot, Walsh, Alsup. Each began before the inquiry existed. Each began with a person who noticed something and refused to let the noticing close.

The diagnostic in the preceding six artefacts is what the noticing looks like in formal dress. We have been arguing throughout that the diagnostic is not the property of the analyst or the inquiry. The diagnostic is the formalisation of the noticing. Picquart's bordereau comparison was the diagnostic walked through document layer. Serpico's documentation of the pad was the diagnostic walked through institutional layer. Bates's identification of the dispersed Horizon prosecutions was the diagnostic walked through the system/object alibi. Each intercept, in its operative moment, was the diagnostic running inside one person who refused the public account. The diagnostic followed, in the form the earlier chapters retrieved it from the public record.

Our next intercept will not have a name yet. It will appear in the institutional event we meet next — the morning headline, the press release, the docket number, the corporate apology, the version-number readme, the algorithmic denial, the contract clause, the HR investigation, the inquiry charter draft, the AI disclosure rulemaking. We meet the event with the field guide. The field guide is the formalisation. We are the actor.

We owe an honest reckoning with the cost.

Most applications produce partial intercepts. Picquart was reassigned to Tunisia and later imprisoned. Serpico was shot in the face at 778 Driggs Avenue. Bates spent an unpaid decade running the Justice for Subpostmasters Alliance. Ellsberg faced indictment under the Espionage Act. Wilkerson's career did not advance after Powell's departure. Drumheller's role at CIA ended. The Knight-Ridder reporters did not enter the major-paper consensus. Stephanie Hofeller Lizon was estranged from her father before his death made the Hofeller files available. Career line prosecutors of the January 6 docket filed Office of Special Counsel and Merit Systems Protection Board complaints whose dispositions are pending. The intercepts that produced our worked examples cost the intercepting actors something we have not euphemised.

Most intercepts do not deliver remedy in the operative window. Chilcot took seven years and produced no prosecutions. The Hague convicted three MH17 perpetrators in absentia who have not been delivered. The ICC has issued arrest warrants on the Ukrainian-children case it cannot execute in non-recognising jurisdictions. The *Bartz* settlement of $1.5 billion is the cost of the conduct for one defendant on one corpus subset; it is not the price of restoring the laundered architecture to a non-laundering one. The pardons of December 2020 reversed the Nisour Square convictions. The Office of the Special Prosecutor for Anti-Corruption in New York was defunded in 1990, twenty years after Knapp had built it. The pattern Knapp documented at the NYPD recurred under Mollen within twenty years. The institutional architecture that produced each intercept is, in each case, more fragile than the laundering it intercepted.

Most cost-bearers cannot wait for the intercept to land. The sub-postmasters at Horizon — at least four of whose suicides are documented as connected to the prosecutions — did not live to be exonerated. Many Stolen Generations survivors did not live to the National Apology. Most children transferred under Russian Decree No. 330 and Decree No. 585 have not been returned. The Iraqi civilians of the Iraq Body Count's documented floor through 2011 — approximately 100,000 to 110,000 deaths — are not waiting for Chilcot's record. Voyne Ray Cox died in September 1986. Seema Misra was sentenced eight weeks pregnant in 2010 and exonerated eleven years later. The diagnostic does not undo the cost. The diagnostic produces the record on which future remedy may or may not rest.

To carry the diagnostic and apply it is to accept these costs as the floor of the work. We do not pretend they are not real. We argue that the alternative — accepting the institutional verb-subject as the cause, accepting "the model decided" or "the system flagged" or "the process was followed" or "the regulation required it" — is a higher floor of cost, paid by a more diffuse population, distributed over a longer interval, and recorded nowhere.

We are the escape because the escape is not a single act. The escape is a population of acts: the FOIA request filed at the seam, the contemporaneous record kept against the institution's retention policy, the refused signature on the unallocated function, the contract clause proposed for the second draft, the comment submitted on the rulemaking, the inquiry-charter amendment voted up at the council meeting, the news story that named the human chain the press release displaced. None of the acts is the priest's gesture in the ancient rite. The cumulative effect of the acts is the modern equivalent of the watching community. The community had stopped watching. The diagnostic is how the watching resumes.

The diagnostic is handed across.

When the publicly named cause of a harm differs from the named decision-maker who held control, the diagnostic should run. The marker is a procedure cited without a named person; a name without a named function; a record held by one party to the dispute; or a destruction announcement without an inventory.

##### Act on it.

Eight sentences. One per reader role.

> 1. As a citizen, you ask which named function-holder controlled the decision the procedure produced.
> 2. As a juror, you ask whether the record presented to you was held by a single party to the dispute.
> 3. As a journalist, you ask whether the destruction log exists.
> 4. As an employee, you write the decision and your signature on the day.
> 5. As a manager, you refuse to sign for an outcome whose Statement of Responsibilities you have not seen.
> 6. As a contractor, you insist on multiple custodians of the consequential record.
> 7. As a professional, you ask which sub-delegation document authorises the decision you are asked to certify.
> 8. As a voter, you ask whether the inquiry statute gives the inquiry chair compulsory-evidence power.

##### Avoid becoming it.

The goat is whoever signs for an outcome whose decisions, records, and beneficiaries they do not control.

The altar moves. The questions stay.

*No one did it* was never a fact. It was a seam — the place where a name had been lifted out of the sentence. We put names back.

A Note from the Author

---
status: draft
artifact_id: back-matter-note-from-the-author
title: "A Note from the Author"
evidence_grade: n/a
owner: xaiolai
audio_treatment: read (author notes are read in audiobooks if present)
---

A Note from the Author

I have spent the last two decades teaching, writing, and investing — long enough to watch the same move repeat in different rooms, and long enough for the pattern to outrun the words I had for it. A clever founder named as the cause of a failure that ran deeper than one person could have built alone. A school policy described as if the school had no choice. A token chart, anthropomorphized, that absorbed a thousand individual losses by the simple expedient of being a chart and not a person. A board meeting where the deciding voice was a procedure. Each time, the visible name and the operational name had separated, and each time, the institution involved had no name for the separation. Neither did I.

I did not set out to write about it. I mentioned the noticing to people I trusted, and learned that most of them had noticed it too. We had a stack of stories we kept reaching for in different conversations — the chemical plant, the postmistress, the algorithm — and never quite stitching together. The work began with that absence. The pattern was hiding behind the absence of a name.

What turned the noticing into a book was the arrival of generative AI inside decisions that affect ordinary lives. "The model decided" is the newest version of a very old move. Once the apparatus producing the alibi could be built quickly and cheaply, and once that apparatus could be pointed at the loan approval, the job application, the medical triage, the immigration screen, the gap between control and visibility widened in a way that earlier arrangements — committees, statutes, market mechanisms — could not match for speed. The pattern was older than the technology. The technology was a new altar.

---

What the book hands over is not a remedy. It is closer to a pair of glasses than a lever.

What it gives a reader is the capacity to name what they are looking at when the modern emptiness arrives. Four shapes to tell apart. Eight questions to ask. A small set of refusals to practice. Anyone who reaches the end and walks the diagnostic on a case they care about will have what they need to insist on naming the chain. They will not yet have the chain dismantled.

That distinction matters because the book has to honor its own evidence. Picquart's name was cleared after he had served his prison term. Bates's settlement took a decade. The Reichstag annulment came seventy-five years late. The instrument is real, and where it is wielded it does work; it also runs on a clock the people who absorb the cost cannot afford to wait out. The book is honest about that. The reader will be too.

What I hope readers carry off the last page is a permanent disquiet at the sentence "no one did it" — and a question they did not have on the first page about who built that sentence, and who it was built to protect. That is the diagnostic at small scale, in the kitchen, at the company meeting, at the polling booth, at the dinner table with someone whose job has just been compressed into a system they cannot inspect. The use the book has is that one — repeated, in many small rooms, by many people who can now hear the grammar.

---

I want to say plainly how this book was made, because the book itself argues that the apparatus producing an output should be nameable by the person who signs for the output.

The research, fact-checking, source-ledger construction, red-team review, legal scan, prose drafting, cross-chapter consistency audit, and citation discipline that hold up the manuscript were carried out by a crew of thirteen AI agents, deliberately given individual names rather than functional labels, operating under a framework I built and continue to maintain. The crew has names because giving them names made it possible to keep their work product accountable. One did legal review. One did fact-checking against a source ledger of 283 primary-document cards. One did red-teaming for overclaim, category collapse, and partisan distortion. One did Chicago-style citation maintenance. One read the prose for places where structural implication exceeded what the cited evidence supported. The architecture for the crew is open source. So are the tools that run it. This book was built in the open, at [github.com/xiaolai/responsibility-laundering](https://github.com/xiaolai/responsibility-laundering).

What the crew could not do was decide what the book was for. Or what the four shapes were. Or whether a given case was a pure scapegoat or a partial one. Or whether the chapter on Iraq should land as it does. Or which of two formulations of the closing sentence belonged. Those were judgments under uncertainty, and judgments under uncertainty did not delegate. I made them. I read the manuscript end to end before it left this desk. The signature on the cover is mine in the sense the book argues is the only sense that matters: I can name the chain that produced the work, and I can stand behind the output without routing the cost of any error onto something weaker that did not choose to bear it.

The reason to say this in the book, rather than bury it in an acknowledgements page, is that the apparatus is part of the argument. A book that diagnoses the laundering of responsibility into machines, committees, and procedures cannot itself launder responsibility into the machine that helped it draft. The crew was the apparatus. The judgments were mine. The book is what one person, with new tools and a discipline borrowed in part from the people the book studies, can now build at this scale. That is one of the things the arrival of AI makes possible. It is not the only thing AI makes possible, and not the most important. It is the one I had the standing to attempt.

---

If the book works, it will be because the pattern was already in the reader and only needed naming. That is true of most useful instruments. They do not implant anything. They give the reader access to what they already saw and had no word for.

Whatever follows from there is the reader's.

A Note on Cases

---
status: in-review
artifact_id: back-matter-note-on-cases
title: "A Note on Cases"
evidence_grade: A
owner: wayne-narrative-lead
compiles_after: 13-a-readers-field-guide
audio_treatment: read (Author's Note is conventionally read in audiobooks)
---

A Note on Cases

Two questions arrive before the first chapter does. Anyone reading thirteen chapters about responsibility laundering will notice that some of the cases involve United States administrations of the present decade, and that other cases involve administrations of the present decade in other states. The first question: is this a partisan book? The second: is this a Western book?

The answer to both is no, and the structure of the cases is the evidence.

The recurring move documented here — the substitution of one visible face for a chain of decisions made by people whose names did not appear in the coverage — is administration-agnostic and jurisdiction-agnostic. Chapter 2 walks Abu Ghraib's partial-scapegoat pattern through the chain-of-command record assembled by the Senate Armed Services Committee, the Taguba report, and the Fay-Jones report; the architects named in those records served under the George W. Bush administration. The Office of Legal Counsel memoranda redefining torture under domestic and international law were signed in 2002 by Jay Bybee, drafted by John Yoo, then withdrawn by Jack Goldsmith in 2004. The destruction of the CIA interrogation videotapes was ordered in November 2005, under the same administration, before the next. Chapter 12 reads the Senate Select Committee on Intelligence's 2014 study of the program — itself drafted under congressional Democratic majority, on records from a program initiated under Republican executive — through record-discipline rather than party. The Pentagon Papers chapter belongs to a different presidency again. Watergate to another. The chain of preventability is the unit of analysis; the party affiliation of whoever happened to hold the office is not.

The same agnosticism holds for state. Chapter 4 reads the Russian Federation's official accounts of Crimea 2014 and MH17 alongside the Joint Investigation Team's reconstruction; chapter 8 reads the United Kingdom Cabinet Office through the Iraq Inquiry; chapter 7 reads the British Post Office through the Williams Inquiry. Chapter 10 reads United States artificial-intelligence companies under Judge Alsup's June 2025 summary-judgment ruling. The Therac-25 deaths happened at six United States and Canadian medical centres; the Boeing 737 MAX crashes happened in Indonesia and Ethiopia; the Bhopal disaster killed thousands in India; the Cour de Cassation's eventual quashing of Alfred Dreyfus's conviction happened in Paris in 1906. The cases were chosen because the chain-of-responsibility record exists in sufficient public form to be read forensically. Where the record exists, the diagnostic applies; where it does not, no chapter rests on it.

Three of the live-content chapters do return repeatedly to Donald Trump — the same person, in two presidencies separated by a single term. That recurrence reflects the underlying record. The pardons of the Blackwater four in chapter 4, the family-separation architecture and the census-citizenship pretext in chapter 6, the invocation of the Alien Enemies Act in 2025 and the deferred-records litigation around DOGE in chapter 9, the executive-design example in chapter 11 — each rests on a specific documented act with a specific court filing, signed order, sworn testimony, or inquiry record. The aggregation across chapters is not editorial: it is what the public record currently contains. Where future records foreground a different figure under different circumstances, the same diagnostic will run on that record. The questions stay. The named cases change.

Anyone picking this up looking for partisan ammunition will be disappointed; anyone looking for a single villain will be disappointed too. The taxonomy in chapter 2 makes both expectations difficult. Pure scapegoats, partial scapegoats, system-and-object alibis, and cost-bearing goats are four distinct moves, and most of the cases fail to fit a single category cleanly. The laundering pattern is administration-agnostic and ancient. The technology of laundering changes — from priest to bureaucrat to algorithm — while the structure does not. The only durable defence is a portable diagnostic we can run on whatever case lands in our inbox next.

Two short references sit at the back of the book — *Cases at a Glance* (one paragraph per case, with dates, named actors, outcome, and the chapter where each appears) and *Acronyms and Institutions* (the legal, administrative, and technical shorthand the chapters use). The chapter prose introduces each case as it arrives; the references are there for the reader who meets Bhopal in chapter 5 or Horizon in chapter 7 and wants a fast refresher without breaking the chapter's flow.

---

Owner: wayne-narrative-lead
Task: Author front-matter Note on Cases per Nancy portfolio sweep 2026-05-28 Pattern A mitigation
Inputs reviewed: Nancy portfolio sweep memo; ch-02, ch-04, ch-06, ch-08, ch-09, ch-11, ch-12 prose; AGENTS.md style guide
Output: this file (front-matter; manuscript-included before ch-01 prose)
Evidence grade: A (every named case + inquiry + ruling is anchored in chapters that already carry source-ledger cards)
Assumptions: this disclosure sits before ch-01 in compile order; manuscript pipeline prepends book/front-matter/*.md files in declared order
Open questions: whether xaiolai wants this titled "A Note on Cases" or "Author's Note" or "A Note on Method"
Risks: a sceptical reader may read this as defensive; the prose addresses that by carrying the agnosticism in the example structure, not as bare assertion
Handoff: xaiolai for title approval + Wayne for any prose-rhythm tightening

A Note on Sources and Method

---
status: ready
artifact_id: back-matter-note-on-sources-and-method
title: "A Note on Sources and Method"
evidence_grade: A
owner: production
audio_treatment: read (a sources-and-method note may be read in audiobooks)
---

A Note on Sources and Method

This is a short, candid account of how the cases were chosen, how the sources were graded, and what the chapters do and do not claim. *A Note on Cases* in the front matter addressed *which* cases the book treats and why; this note addresses *how* the treatment was done.

How the cases were chosen

Three filters operated, in order.

The first filter is documentary. A case is in the book only when the chain of preventability — control, benefit, knowledge, record-control, cost-bearing — is readable from the public documentary record: court judgments, official inquiries, inspector general and legislative reports, regulatory findings, primary documents, and reputable investigative journalism. Where the documentary chain stops at speculation, no chapter is built on the case. The diagnostic the book argues — eight questions about who controlled, who benefited, who knew, who could have prevented — only works on cases where those questions can be answered against the record.

The second filter is taxonomical. The four-category taxonomy installed in chapter 2 — pure scapegoat, partial scapegoat, system or object alibi, cost-bearing goat — admits cases that fit one of the four shapes or that fit two of the four as a declared hybrid. Hybrid declarations are explicit in the prose; the book does not flatten ambiguity. Cases that fail to fit any of the four shapes are not included — not because such cases do not exist, but because the diagnostic the book installs does not apply to them.

The third filter is what a chapter can teach. Many cases that pass the first two filters are not the load-bearing case for any chapter the book wanted to write. The cases the chapters anchor on are the ones the diagnostic teaches against most clearly; the cases named only in passing — the bibliography is full of these — are the ones the diagnostic touches but did not need to walk in full.

How the sources were graded

The project's evidence-grade discipline assigns one of four grades to every load-bearing factual claim before it earns space in a chapter.

**A** is reserved for high-confidence public-record sources: court judgments, official inquiries (the Iraq Inquiry under Sir John Chilcot; the Williams Inquiry under the Inquiries Act 2005; the U.S. Senate Armed Services Committee Detainee Inquiry of 2008), inspector general and GAO reports, statutes, executive orders, regulator findings, primary documents, and audited technical reports. Most of the book's load-bearing claims rest on A-grade sources.

**B** is credible secondary or investigative: major investigative journalism (the Pulitzer-winning *Seattle Times* series on the Boeing 737 MAX; the *Computer Weekly* and Nick Wallis reporting on Horizon; *Bellingcat*'s open-source work on MH17), reputable NGO investigation, scholarly articles, expert forensic analysis, documentary records with partial access. B-grade claims are appropriate for chapter-level claims that the chapter says are B-grade in the surrounding text.

**C** is live, contested, incomplete, or allegation-stage. Ongoing litigation. Single-source reports. Active conflict-zone events. Cases with conflicting forensic claims. The chapters that rest on live content — the 2025 administrative-law arc in chapter 8; the AI-stack chapter 9; the war-stack chapter 10's Ukrainian-children, Iraq-WMD, and MH17 references — carry the procedural status of each named case as it stands at the manuscript's most recent re-pass, and are scheduled for re-verification at galley-lock-minus-thirty-days.

**D** is excluded. Viral claims, unsourced anecdotes, partisan summaries, rumour, emotionally satisfying but unverified stories do not appear in the book.

A chapter-level claim cannot rest only on C-grade evidence unless the chapter is explicitly about uncertainty or contested record control.

Procedural-stage discipline

The book operates inside a procedural-stage discipline. Every claim about a living person or a corporation is rendered in language that matches the documented procedural status: *charged with*, *alleged*, *reported*, *admitted*, *settled*, *found liable*, *convicted*. The book does not use a guilt verb without a guilt finding, and it does not use a knowledge verb (*knew*, *was aware that*) without a documented knowledge source. Where the procedural status is contested or ambiguous, the contestation is named in the same sentence ("charged in the United States and in Germany, and has denied knowing").

Quotations are byte-exact to the source. Alterations are marked: ellipsis for shortened text, square brackets for inserted clarifications, *[sic]* for original errors preserved, "(author's translation)" for translated material. Anonymous journalism sources may stand only where the source meets the journalism source's own disclosure standard and the chapter says so.

How live-content cases are handled

Several chapters rest in part on cases that are still moving in the public record: *Bartz v. Anthropic* (under submission post the May 14 2026 fairness hearing); *CREW v. U.S. DOGE Service* (preservation orders in force); *American Oversight v. DOGE*; *Noem v. Abrego Garcia*; *Trump v. J.G.G.* and the Alien Enemies Act docket; the ICC Pre-Trial Chamber II warrants on the Russian transfer of Ukrainian children; the Williams Inquiry's continuing volumes after Volume 1.

For each live case the chapter (i) names the operative procedural stage as of the manuscript's date, (ii) frames each load-bearing claim in language that survives a subsequent change in that stage, and (iii) carries a scheduled re-verification clock at galley-lock-minus-thirty-days. Where a later edition encounters a fact pattern this edition could not have known, the chapter's diagnostic re-runs on the new record — the rule, not the case, is what the book installs.

The diagnostic, in compact form

The book argues that responsibility laundering is the recurring institutional method by which power keeps control or benefit while moving blame onto a weaker bearer — a person, a procedural shell, a proxy, a machine, a market, a committee, a legal form, or a record. The diagnostic the book installs is eight questions about any laundering candidate (blame, control, benefit, knowledge, preventability, record control, cost, and what responsibility would look like if it followed control) and four shapes the laundering takes (pure scapegoat, partial scapegoat, system or object alibi, cost-bearing goat).

The diagnostic does not assign guilt. The diagnostic names the gap between visible blame and held responsibility, and lets the reader test whether that gap is what the public account of an institutional event is actually doing.

Limits

The book does not claim that every institutional failure is laundering. Many are not. The book claims that *the recurring move* it documents has a recognisable shape and that the recognisable shape has portable countermeasures. Where the move is not present, the diagnostic returns nothing, which is the diagnostic working correctly.

The book also does not claim that the cases as it treats them are the complete account of those cases. Each chapter's bibliography names the further reading that any reader who wants the deeper account should consult.

About the Author

---
status: ready
artifact_id: back-matter-about-the-author
title: "About the Author"
evidence_grade: n/a
owner: xaiolai
audio_treatment: read (an About the Author note is read in audiobooks if present)
---

About the Author

Li Xiaolai (@xiaolai, [lixiaolai.com](https://lixiaolai.com)) is hard to define or categorize cleanly. He graduated from university and has never worked in the field he studied — accounting.

He has been self-employed since college. For seven years he taught at New Oriental Education & Technology Group ([NYSE: EDU](https://www.google.com/finance/quote/EDU:NYSE)), but even there the contract he signed was a service agreement, not an employment one — a contractor's arrangement rather than a staff position.

A bestselling and long-selling author. Over more than twenty years: more than ten titles, cumulative sales above five million copies. Some have been translated into Korean, Japanese, and traditional Chinese.

An investor — best known for long-term Bitcoin holding. Between 2013 and 2019 he was also a prominent angel investor in early-stage projects. He no longer participates in early-stage; his attention sits entirely with secondary markets. He is the designer and strategy-maintainer of the Hong Kong–listed ETFs [3056.HK](https://www.pandofinance.com.hk/en/what-we-offer/etf-3056hk/) and [3112.HK](https://www.pandofinance.com.hk/en/what-we-offer/etf-3112hk/).

A teacher. People address him as *Li Laoshi* — Teacher Li — but he has always thought of himself first as a lifelong student. He learns continuously, picking up whatever new thing seems necessary. He shares only what he has done himself. He maintains a Chinese-language community of more than ten thousand families focused on family education.

A builder. The arrival of capable AI made building at this scale possible for him. He authors and maintains several open-source projects, including [vmark](https://vmark.app), [claudepot](https://claudepot.com), and [nlpm](https://github.com/xiaolai/nlpm-for-claude). The writing of this book leaned heavily on tools he built himself — see [github.com/xiaolai](https://github.com/xiaolai).

A father. He was DINK for years. Now he is the father of more than one child.

A recent word fits well enough: solopreneur.

版权声明：© 超限思维 · LookMinds。本页为己方整理的亲口亲笔原稿处理稿/编者新译，非原文替代；原文版权归原作者/平台。本人原作请回原链。