Comparative officer liability — the primary provisions

The root README’s question ladder answers four objections with facts drawn from outside American law: that officer liability is foreign to the American tradition, that it would cost the country its lead, and that no other country imposes it on its own industry. This file holds the sources behind those answers.

What this file is not. It is not part of the Act’s authority base. The statute cites none of it, and the table of authorities is scoped to the statute and companion by its own terms. These are the underlying provisions: material used to answer objections, pinned here so that a reader who doubts a sentence on the front page can check it without writing to ask.

Status. Compiled 20 August 2026. Foreign statutes are quoted from official or first-party publications, which control; translations are the publishers’ own. Two items are marked open at the end, and are open on purpose.

1. China

PRC Criminal Law art. 31.

Answers: “How does this proposal compare with officer-liability regimes in other jurisdictions?”

The proposition. Chinese criminal law already imposes dual punishment — the entity fined, the responsible individuals criminally punished — generally across the code, and has since the 1997 revision. Officer liability is not an American import into China; it is the ordinary law of the other large economy, on a broader class of conduct than this Act reaches.

The text, as published in English by the Supreme People’s Procuratorate of the PRC:

Article 31 An entity shall be fined for the crime it has committed. The directly responsible persons in charge and other directly responsible persons shall also be subjected to criminal punishment. Where the Special Provisions of this Law or other laws provide otherwise, those provisions shall prevail.

Source: Supreme People’s Procuratorate, English translation of the Criminal Law of the People’s Republic of China (en.spp.gov.cn), retrieved 20 August 2026. The National People’s Congress English text (npc.gov.cn) renders the same article as: “Where a unit commits a crime, it shall be fined, and the persons who are directly in charge and the other persons who are directly responsible for the crime shall be given criminal punishment.” The Chinese original controls.

Provenance of the date. The Law was adopted 1 July 1979 and revised at the Fifth Session of the Eighth National People’s Congress on 14 March 1997; art. 31 sits in Chapter II, Section 4 (“Crimes Committed by a Unit”), added at that revision. Subsequent amendments I through XII have not disturbed it.

Scope note, stated so the comparison stays honest. Article 31 is a general attribution principle applying wherever the Specific Provisions define an offense as capable of commission by a unit—a wide but not universal field. The comparison is one of legal breadth: Article 31 operates across many areas of the economy, while this Act is confined to defined frontier-AI conduct and imposes individual duties only on natural persons satisfying its controlling-person test. Neither comparison establishes a fixed headcount or implies that every organization or officer is liable.

2. Germany

§ 130 OWiG, with § 9 and § 30.

Answers: “No other country does this to its own industry.”

The proposition. German law reaches the individual for a failure of supervision that enabled a contravention, and reaches managing directors and board members specifically, not merely the nominal proprietor. It is weaker than this Act and it is an administrative offense rather than a crime — both stated on the front page, because overstating it would be the easier claim and the falsifiable one.

The text (Gesetz über Ordnungswidrigkeiten § 130(3), first sentence):

Die Ordnungswidrigkeit kann, wenn die Pflichtverletzung mit Strafe bedroht ist, mit einer Geldbuße bis zu einer Million Euro geahndet werden.

Source: gesetze-im-internet.de, the Federal Ministry of Justice’s official publication, retrieved 20 August 2026. Fassung per the Eighth Act amending the Act against Restraints of Competition, 26 June 2013 (BGBl. I S. 1738).

The structure, in three provisions. § 130 makes it an Ordnungswidrigkeit for the owner of a business or undertaking to omit the supervisory measures required to prevent contraventions of duties incumbent on the owner, where the contravention would have been prevented or made substantially more difficult by proper supervision — the measures named in the text including appointment, careful selection, and monitoring of supervisory personnel. § 9 (Handeln für einen anderen) extends the addressee to organs and authorized representatives, which is what brings a GmbH’s Geschäftsführer and an AG’s Vorstandsmitglieder inside it. § 30 is the entity-side fine, to which § 130 is the usual predicate where no individual contravention can be proved.

Why the €1 million figure and not the larger ones. § 130(3) sets one million euros where the underlying breach of duty carries a criminal penalty. Larger ceilings quoted in practitioner literature — ten and five million under § 30, and forty and twenty million in the sanctions context — are the entity’s exposure, not the supervisory offense’s. The front page quotes the § 130 figure because the comparison being drawn is about the duty reaching a named person.

Scope note. Germany has no general corporate criminal code; a draft Verbandssanktionengesetz has been proposed and not enacted. The front page says so.

3. The United States, 1879 to 1892

The pork war and the export-inspection acts.

Answers: “Won’t this hand the lead to someone else?”

The proposition. The United States met a ten-nation safety embargo with inspection law carrying criminal penalties, and its export trade recovered and grew. The claim the front page makes is deliberately narrower than the story invites, for reasons at the foot of this section.

The embargo. From 1879, Germany and nine other European states — Austria-Hungary, Denmark, France, Greece, Italy, Portugal, Romania, Spain, and the Ottoman Empire — banned American pork wholly or partly, citing trichinosis. American meat exports fell from about $134 million in 1881 to about $69 million in 1882. Sources: Newman, Playing the Defense: The Beef Trust, Cronyism, and the 1891 and 1906 Meat Inspection Acts, 29 Indep. Rev. (2024), at 31, citing Young (1989) and Historical Statistics of the United States (2006), vol. 5, 551; and the Wikipedia article “Pork war” for the roster of banning states.

The answer. An Act of 30 August 1890 (26 Stat. 414) required inspection of salted pork and bacon intended for export; an Act of 3 March 1891 (26 Stat. 1089) extended inspection and certification to live cattle, hogs, and their carcasses and products. Sources: USDA Food Safety and Inspection Service, “Our History” (fsis.usda.gov); National Research Council, Cattle Inspection (1990), historical review chapter, NCBI NBK235649.

The result, from the primary source. President Harrison’s annual message to Congress, December 1892, reporting on the first year of operation:

a comparison of the four months of May, June, July, and August, 1892, with the same months of 1891 shows an increase in the number of pounds of our export of pork products of 62 per cent, and an increase in value of 66½ per cent

The same message reports dressed-beef exports rising from 137,900,000 lbs in 1889 to 220,500,000 lbs in 1892, and 394,607 head of live cattle exported in the past year against 205,786 in 1889. Source: Message of the President, digitized by the University of Wisconsin Libraries (search.library.wisc.edu), retrieved 20 August 2026.

The half that hurts, and why the front page carries it. Inspection alone did not reopen Germany. The German ban lifted in 1891 after President Harrison threatened a retaliatory embargo on German sugar beets under the McKinley Tariff’s retaliation provision, and other states followed. Any version of this argument that runs bans → inspection → exports up 62 per cent asserts a causal chain the record does not support, and a hostile reader finds the sugar beets in one search. The front page therefore claims only that verifiable safety was necessary rather than sufficient — the asset the diplomacy had to hold before it could be spent, and the thing that kept the market open afterward.

One claim declined. A dispatch from an American diplomat in Copenhagen, 1891, to the effect that rigid inspection would return more than its cost, was drafted into an earlier version of this answer and cut: it could not be pinned to a primary source in the session that used it. If it is ever located in Foreign Relations of the United States, it may return; until then it appears nowhere.

4. The United Kingdom

Health and Safety at Work etc. Act 1974, s. 37, and the consent-connivance-neglect family; separately, Financial Services and Markets Act 2000, ss. 60(2A)–(2B) and 66A.

Answers: “No other country does this to its own industry” — and answers it more strongly than Germany does, because the British provision is criminal.

The proposition. British law reaches the individual director or manager criminally, on a negligence floor, through a formula that recurs across the statute book and has done since 1974. Two design choices this Act had to argue for at length are already enacted in a single English sentence: liability may rest on neglect rather than knowledge, and it attaches to a person by what they were doing rather than by their title.

The honest limit, stated first because overstating it would be the easier claim and the falsifiable one: s. 37 is derivative. It creates no duty of its own. It attaches an individual to an offense the body corporate has already committed under some other provision. It is therefore not a British version of this Act, which creates its own duties and then attaches persons to them. What s. 37 establishes is narrower and still useful: the mechanism — a named human answering criminally for a corporate failure they had power to prevent — is ordinary British legislative furniture, not an American peculiarity requiring eighty years of doctrine to justify.

The text (Health and Safety at Work etc. Act 1974, s. 37):

(1) Where an offence under any of the relevant statutory provisions committed by a body corporate is proved to have been committed with the consent or connivance of, or to have been attributable to any neglect on the part of, any director, manager, secretary or other similar officer of the body corporate or a person who was purporting to act in any such capacity, he as well as the body corporate shall be guilty of that offence and shall be liable to be proceeded against and punished accordingly.

(2) Where the affairs of a body corporate are managed by its members, the preceding subsection shall apply in relation to the acts and defaults of a member in connection with his functions of management as if he were a director of the body corporate.

Source: legislation.gov.uk, official revised text, retrieved 21 August 2026.

The two clauses that do this Act’s work.

“Attributable to any neglect.” The predicate is not knowledge, intention, or recklessness. Negligence suffices for individual criminal liability. That is the floor SEC. 6(c) sets for custodial exposure, and in the United States it had to be defended against the modern scienter presumption — Staples, Rehaif, Ruan. In the United Kingdom it is the 1974 baseline.

“Or a person who was purporting to act in any such capacity.” Substance controls over title, on the face of the statute, fifty-two years before this draft. SEC. 4 spends a section arriving where s. 37 arrives in eleven words.

The family, and the split inside it that matters. The construction recurs across the statute book, but not uniformly — and the variation tracks the same axis this Act uses to separate its own tiers. Each provision below was retrieved from legislation.gov.uk on 21 August 2026.

With the neglect limb — regulatory and public-welfare statutes:

Environmental Protection Act 1990, s. 157, in wording identical to s. 37: “committed with the consent or connivance of, or to have been attributable to any neglect on the part of, any director, manager, secretary or other similar officer… or a person who was purporting to act in any such capacity.”
Data Protection Act 2018, s. 198(1)(b): “with the consent or connivance of or to be attributable to neglect on the part of” a director or similar officer; s. 198(4)(b) applies the same three pathways to partners in Scottish partnerships.

Without it — dishonesty offenses:

Fraud Act 2006, s. 12: “If the offence is proved to have been committed with the consent or connivance of— (a) a director, manager, secretary or other similar officer of the body corporate, or (b) a person who was purporting to act in any such capacity, he (as well as the body corporate) is guilty of the offence.” Consent or connivance only. No neglect limb.
Bribery Act 2010, s. 14: consent or connivance of “a senior officer of the body corporate or Scottish partnership.” No neglect limb, and a narrower class of person.

What that split is worth. Parliament does not extend negligence liability to individuals indiscriminately. It attaches the neglect limb where the offense is public-welfare regulation — safety at work, the environment, data — and withholds it where the offense is dishonesty. That is precisely the line this Act draws between SEC. 6(a), the due-care offense with a negligence floor, and SEC. 6(b), the felony tier requiring knowing or willful conduct. The British statute book already sorts these provisions on the same axis the Act sorts its tiers on. For the criminal-law seat, that is a stronger comparative point than the bare existence of officer liability, and it was found by checking four provisions that were assumed to be identical and are not.

The adjacent regime, structurally closer to SEC. 8 than anything else in this file. The Senior Managers and Certification Regime, in authorized financial services firms, does two things this Act does and Park does not.

Responsibility is allocated in advance, in writing. FSMA 2000, s. 60(2A): where an application seeks approval to perform a designated senior management function, the regulator “must require the application to contain, or be accompanied by, a statement setting out the aspects of the affairs of the authorised person concerned which it is intended that the person will be responsible for managing in performing the function.” s. 60(2B): “A statement provided under subsection (2A) is known as a ‘statement of responsibilities’.”

And the individual answers personally for failing to prevent. FSMA 2000, s. 66A(5)(d) provides that a senior manager is guilty of misconduct where “the senior manager did not take such steps as a person in the senior manager’s position could reasonably be expected to take to avoid the contravention occurring (or continuing).”

Source: legislation.gov.uk, retrieved 21 August 2026.

Stated plainly, because it would be easy to overstate: this is regulatory, not criminal. s. 66A defines misconduct for the purposes of FCA enforcement under s. 66. It is a disciplinary jurisdiction over approved persons, not a prosecution. This file says so rather than blurring it, exactly as it does for § 130 OWiG.

Its relevance is architectural. Park asks, after the harm, who stood in responsible relation to it. SMCR asks before: who is answerable for this, in writing, filed with the regulator, and did that person take reasonable steps. SEC. 8’s certification is the same instinct given a criminal consequence. The point for a skeptical reader is that the British state already accepts the premise that a named human being should be on record as answerable for a risk before it materializes — in banking, where the harm is money.

Scope note, and why Britain is not the venue. The United Kingdom has enacted no frontier AI statute; its approach has been deliberately non-statutory. Two consequences follow. There is nothing for a British SEC. 3(c)(4) to borrow, because no enacted framework exists to freeze. And there is no sub-national entry point: the “one state is enough” theory and the biometric-privacy precedent behind it have no British equivalent, the subject belonging to Westminster. Britain is better ground doctrinally and worse ground strategically, and this file records both.

5. The United States, today — the codified officer

Added 23 August 2026. The sections above trace the doctrine abroad and in history. This one records something nearer and stranger to miss: Congress has twice written the phrase “responsible corporate officer” into the United States Code itself — not doctrine applied by courts, but the words, enacted, in the criminal definition of “person.” Both provisions retrieved verbatim 23 Aug 2026 (uscode.house.gov / LII; ⚠ R under the census rubric); the scholarly apparatus is from two sources held on the shelf.

The Clean Water Act. 33 U.S.C. § 1319(c)(6): “For the purpose of this subsection, the term ‘person’ means, in addition to the definition contained in section 1362(5) of this title, any responsible corporate officer.” The subsection it serves is the criminal one, and its offenses are felonies with a “knowing” element — a more aggressive structure than Dotterweich’s and Park’s misdemeanours. The Ninth Circuit reads Congress’s retention of the phrase after Park as ratification of the doctrine (United States v. Iverson, 162 F.3d 1015, 1023–24 (9th Cir. 1998), quoting United States v. Brittain, 931 F.2d 1413, 1419 (10th Cir. 1991)); convictions of officers under it have been affirmed across circuits (Iverson; United States v. Hansen, 262 F.3d 1217 (11th Cir. 2001); United States v. Ming Hong, 242 F.3d 528 (4th Cir. 2001)). And the doctrine has been extended to civil penalties: “the rationale for holding corporate officers criminally responsible … is even more persuasive where only civil liability is involved” (United States v. Hodges X-Ray, Inc., 759 F.2d 557, 561 (6th Cir. 1985), imputed to the CWA by district courts since).

The Clean Air Act. 42 U.S.C. § 7413(c)(6): “For the purpose of this subsection, the term ‘person’ includes, in addition to the entities referred to in section 7602(e) of this title, any responsible corporate officer.” Added in 1977 — after Park — which courts likewise treat as congressional approval of the doctrine. (The 2025 federal AI roadmap cites this same statute — as permitting paperwork to streamline for data centers: two visions.)

The scholarship, in hand. Lyness, Revitalizing the State Environmental Responsible Corporate Officer Doctrine, 64 B.C. L. Rev. 253 (2023) — the survey this section draws on (its Part II read; its Part III catalogues the state environmental officer doctrines, ten states, the groundwork for for legislators § 4) — describes the doctrine as “a potent tool for federal environmental enforcement for over half a century,” and its revitalization agenda for the states runs: individual civil liability; liability without a mens rea requirement; uniformity; broad enforceability. Those four goals are this Act’s own design list — SEC. 6 and SEC. 10(b); the public-welfare classification of SEC. 1(a); the model-act form; SEC. 10’s remedy set. Beside it: Hustis & Gotanda, The Responsible Corporate Officer: Designated Felon or Legal Fiction?, 25 Loy. U. Chi. L.J. 169 (1994) — practitioners writing while the enforcement wave crested: officers and managers were 80% of all individuals prosecuted for environmental crimes since 1983, under an explicit EPA/DOJ policy of charging “the highest ranking officers … for whom personal culpability can be shown.” Its title survives as the objection’s honest form; its first pages record that the felony question was settled by enactment.

What this section does for the Act. The strongest form of the “extraordinary doctrine” objection assumes the doctrine lives only in two old cases about drugs and eggs. It does not. It lives in the Code, twice, as felony architecture, ratified after Park, extended to civil penalties, enforced for fifty years — in the field whose hazards (diffuse, delayed, catastrophic, institutional) most resemble this one’s. The Act’s innovation is the subject matter, not the mechanism.

§ 5 addendum, same night — a precision owed, and three finds from the shelf’s long documents

The precision (candidate erratum, maintainer to number). The paragraph above maps Lyness’s four-goal agenda onto the Act’s own provisions. His Part IV, read in full after that paragraph was committed, sharpens the fit and the honest statement is this: on three of the four goals — no mens-rea requirement (in the civil form), uniform application, broad enforceability — the convergence holds. On the first, it does not: Lyness argues for individual civil liability and “only civil liability,” expressly excluding the criminal form as the target of the “extreme and illegitimate” critiques (64 B.C. L. Rev. at 297–99). The Act takes the criminal road with a due-care culpability floor (SEC. 6(c)) — the design he would critique, answered at the case and known objections, not assumed away. His fourth goal also names a mechanism the Act deliberately lacks: citizen suits. The Act’s route for private information is the SEC. 11 award (private knowledge funding public enforcement), not a private action; a reader pressing the citizen-suit question should be pointed at that design choice, which bracketed matter does not currently discuss and a drafting session may wish to.

§ 5 addendum, 25 August — the full article read, and two harder points it makes. The complete PDF is now held (SSRN 4186172; the reading note is in the private library). The civil-only finding above stands unchanged and needs no further correction. Two further passages, quoted rather than paraphrased because they cut at the Act’s foundations, belong on this page so no reviewer finds them first:

The empirical claim about state legislatures. “Indeed, state legislatures have largely chosen not to include responsible corporate officer liability within the criminal provisions of environmental laws” (64 B.C. L. Rev. at 297–98). This project’s census counts what states have enacted for AI; Lyness counts what they declined to enact in the doctrine’s home field. A sponsor asking “has any state done this criminally” is entitled to that answer, and it is not a comfortable one. The Act’s response is not that he is wrong but that the risk profile differs: environmental harm is largely remediable and insurable, and the Act’s own argument for the criminal road rests on harms that are neither. That answer is asserted here, not proven, and the comparative seat should test it.
The misdemeanor point, which bears on the Act’s own lineage. “To be sure, both Dotterweich and Park involved criminal liability. But both involved misdemeanor criminal liability during a time when the immediate and collateral consequences were different” (at 298). The Act is named for the doctrine those cases created and cites them as its foundation. If the authority they supply is misdemeanor-scale, then the Act’s penalty architecture reaches beyond what its own precedents establish, and the gap must be argued for rather than inherited. This is now an open question for the criminal-law and proportionality seats jointly, and it is the strongest doctrinal objection in the scholarship this project relies on.

Neither point is an erratum: nothing published was inaccurate. Both are disagreements from a source the Act uses, recorded at full strength, which is what this page is for.

Find one — the doctrine reaches inside government. United States v. Dee, 912 F.2d 741, 747–49 (4th Cir. 1990), cert. denied, 499 U.S. 919 (1991): civilian U.S. Army managers of a federal chemical facility, convicted under RCRA, their culpability arising “solely from their ongoing failure to comply … during the period they were responsible” (as summarized at Hustis & Gotanda, 25 Loy. U. Chi. L.J. at 172 n.15, article in hand). The responsible-officer principle is not confined to the private sector — a fact with obvious weight in a record whose most-documented incident occurred at a government institute.

Find two — the enforcement wave, quantified. The same article’s opening pages record a DOJ official’s estimates from the early-1990s wave: officers and managers were 80% of individuals prosecuted for environmental crimes since 1983, and 68% of all jail time ever served for environmental violations was imposed in that period (at 169–70 & n.4). And the modern counterpoint, from Lyness’s Part IV: by FY2020 federal environmental enforcement had fallen to its lowest levels in twenty years — 82 civil judicial cases concluded, 81 criminal referrals (at 295). The pairing is the state-tool argument in two numbers: the doctrine works when enforced, and federal enforcement is not a constant anyone may assume.

Find three — the case studies, and a fifth venue. Lyness’s Part V (read in full) offers Massachusetts and Oregon as adoption models: Massachusetts as the “ideal jurisdiction” — near-uniform environmental definitions of “person” that already include “any officer, employee, or agent of said person” (Mass. Gen. Laws ch. 21C, § 2), so that “[a]ll it would take is for one advocate to convince one judge” (at 303–05); Oregon as the case where piecemeal definitions mean “legislative adoption would be more comprehensive. A single statute could create a uniform category” (at 305–06). The second sentence is the model-act argument, stated by the doctrine’s own scholar; the first adds Massachusetts to the venue map — see for legislators §§ 4–5.

Open items in this file

The Copenhagen dispatch (§ 3), unpinned and therefore unused.
PRC Criminal Law Amendment XII (adopted December 2023, effective 1 March 2024) raised penalties for certain breach-of-duty and bribery offenses by personnel of private enterprises. An earlier draft of the § 1 answer stated more broadly that several maxima rose from five years to ten; the wider claim was not verified and was cut. The narrower, verifiable version may be worth pinning if the comparison ever needs it.

Corrections to FrontierAIAccountabilityProject@proton.me; they enter the errata register like everything else. Nothing in this file is legal advice, and nothing in it is cited by the statute.