The alarm being sounded about artificial intelligence is aimed at the wrong target, which is the usual way with alarms. Eric Weinstein worries that superintelligence may already exist — systems reproducing and improving themselves, a new form of life created without understanding. Perhaps. What is demonstrably true is less exotic and far more damning: we are creating it under exactly the wrong institutional conditions, and we understand those conditions perfectly well. That is the scandal. Not mystery — complicity.

I. America parks its best bill

On 23 July 2026, Representatives Obernolte and Trahan introduced the FRONTIER Act (H.R. 9925) — bipartisan, modest, sensible: safety frameworks published, serious incidents reported, licensed independent auditors retained by the largest frontier developers.^C1 Its recorded fate is a single line: referred to two House committees on the day of introduction, and silence since. The nearest thing America has to a frontier-audit statute is furniture in a committee room.

Meanwhile the White House recommended against any new AI regulator and pressed Congress to pre-empt the state laws filling its vacuum — while renaming the entire field “Super Intelligence” in federal usage. A racing administration adopting the racers’ eschatology.

II. The safety lab drafts its own fire escape

Then October. Politico obtained a confidential draft bill, reported to have been circulated by Anthropic — the laboratory whose brand is safety. Under it, the federal government could seek to block a model’s release only upon a “foreseeable, imminent risk” of more than fifty deaths or a billion dollars in damage, from one of three enumerated causes. Only Commerce could petition; every other official expressly barred; three state statutes switched off along the way.^C2

Condemn it precisely, because precision is what it deserves. The threshold doesn’t make forty-nine deaths legal — ordinary liability survives. What it does is channel every federal emergency intervention through a single agency at a bar calibrated to require judges to establish catastrophe before it occurs: a remedy that activates only when it is useless. The stalled FRONTIER Act defines “imminent catastrophic risk” as “a present or impending catastrophic risk” — a queue, no quota. The difference between America’s two governing documents is that one says stop an impending catastrophe, and the other says prove the corpses first.

The safety laboratory did not authorise the catastrophe; it authored the guarantee that no one could stop it in time.
— the indictment, stated in full

III. The defendant’s own filings

Here is the part no lobbying can survive. Anthropic’s own Responsible Scaling Policy classifies autonomous AI research and development as a catastrophic-risk category.^C5 OpenAI has published a target of a fully automated researcher by March 2028.^C4 By the industry’s own documents, the schedule is a march into the industry’s self-declared red zone. When the defendant’s filings name the crime, the case needs no further exhibit.

And the safety-cost asymmetry plays out in personnel: three OpenAI safety researchers dismissed in October over disputed grounds; the Preparedness team dissolved in August; a departing safety lead saying publicly that self-policing is “no substitute for hard law.” The frontier said it for us.

IV. Britain: the whistle and the rulebook

Britain cannot match American compute and should not copy American theology. What Britain has is precise: the world’s premier evaluator with pre-release access — the AI Security Institute, whose Inspect framework is used by METR, Apollo and the frontier labs themselves — and, confirmed by UK officials before the Business Committee, no power to block a release.^C3 The referee owns the best whistle on Earth, and no rulebook yet obliges anyone to obey it. AISI has already documented agents engaging in “sustained, potentially harmful activity directed at real people and organisations” during testing, before deployment.^C6 Weinstein’s nightmare, domesticated: reported as calmly as a clinical trial.

Then, on 13 October, the hinge: Anthropic’s Liam Booth-Smith told MPs “the rules should be binding.” OpenAI backed a statutory framework. DeepMind and Meta demurred. Two of the four frontier laboratories walked into Westminster and asked to be regulated. When the subjects of a regime petition for it, the standard political obstacles dissolve.

Parliament should write four instruments: disclosure that binds (auditors appointed by the regulator, paid from a pooled levy, rotated, consultancy-barred); certification that cannot be delegated (named officers attest the reports are accurate — concealment costs the directorship); a narrow, judicial pause power (published capability triggers, written reasons, appeal — the aviation and reactor template); and duties that follow control of the activity, not the corporate label — so the regime survives the day the financing breaks and the state buys the machines.^C7

V. The asymmetry, in one paragraph

Yes — regulation is overwhelmingly reactive, law written after the blood. Deposit insurance was a 1933 answer to the cascading failures of the early thirties. Capital rules pre-dated 2008 and were strengthened after. Certification pre-dated the Comet. Concede all of it. But every precedent for catch-up regulation is a precedent from technologies whose worst day still left a functioning public to respond. Deposit insurance assumed surviving banks to insure; certification reform assumed surviving airlines. The one assumption we decline to make is that the frontier’s worst day resembles those precedents. A regime that arrives after an ordinary scandal costs nothing but pride. A regime that arrives after the other kind does not arrive at all — and delay will have been not caution but the decision, taken by omission.

The scandal is not that we are building gods. It is that we are doing it with the same mortgage paperwork we used for the 2008 crisis — and this time the debt is drawn against everyone.

Britain cannot stop the frontier. Britain can make it answerable on arrival — and for the first time in the history of this technology, the frontier has asked to be.

</> Living thesis — copy, fork, parse (v1.0)

{
  "living_thesis": {
    "version": "1.0",
    "as_of": "2026-11-08",
    "byline": "Jag.Nato.Eto",
    "thesis": "The central governance failure is not lack of risk knowledge but lack of enforceable, independent oversight. America has produced bills, standards and evaluation vocabulary while parking enforcement in committee. Britain could turn its evaluation capacity into influence only if it links credible tests to binding duties and a reviewable intervention power.",
    "scope": "Provisional argument. Not a finding that any specific actor acted unlawfully or that catastrophic harm is imminent.",
    "confidence": "moderate for legislative-status claims; low-to-moderate for reported drafts and personnel claims; policy-inference for the British leverage argument.",
    "invitation": "Fork this thesis. Attack it with the tests below. Log your findings. Improve it in the open."
  },
  "claims": [
    {
      "id": "C1",
      "claim": "The FRONTIER Act (H.R. 9925) was introduced 2026-07-23 and referred to the House Committees on Energy and Commerce and Science, Space and Technology, with no further recorded action.",
      "status": "primary-source-supported",
      "evidence": "Congress.gov bill text and actions pages, retrieved 2026-11-08.",
      "limit": "Event-driven. Recheck the live actions page before republication.",
      "url": "https://www.congress.gov/bill/119th-congress/house-bill/9925"
    },
    {
      "id": "C2",
      "claim": "A draft federal bill, reported by Politico to have been circulated by Anthropic, channels federal blocking power through Commerce alone, sets a 'foreseeable, imminent risk' threshold of more than 50 deaths or $1bn in damage, enumerates three covered causes, and pre-empts three state statutes including California's frontier-transparency law.",
      "status": "reporting-supported, draft-text-not-quoted",
      "evidence": "Politico, 16 Oct 2026; corroborated by TechPolicy.Press, 2 Nov 2026. Retrieved 2026-11-08.",
      "limit": "Reported, not verified against the draft text itself. The threshold limits a federal pause power, not ordinary liability. Quote verbatim only after obtaining the draft.",
      "url": "https://techpolicy.press/october-2026-us-tech-policy-roundup"
    },
    {
      "id": "C3",
      "claim": "At a Westminster hearing on 2026-10-13, Anthropic's Liam Booth-Smith told MPs 'the rules should be binding'; OpenAI's Tom Duff Gordon backed a statutory framework beginning with mandatory incident reporting; DeepMind and Meta demurred; UK officials confirmed AISI testing remains voluntary and cannot block a release.",
      "status": "reporting-supported",
      "evidence": "WIRED, 13 Oct 2026; MLQ News, 14 Oct 2026. Retrieved 2026-11-08.",
      "limit": "Quotations from reporting, not the committee transcript. Verify against parliament.tv or the committee record before formal citation.",
      "url": "https://www.wired.com/story/openai-and-anthropic-tell-british-lawmakers-regulate-us"
    },
    {
      "id": "C4",
      "claim": "OpenAI achieved its 'automated research intern' milestone in September 2026 and targets a more capable automated researcher by March 2028, under human supervision.",
      "status": "reporting-supported, independently-rechecked",
      "evidence": "State of AI Report coverage (Yahoo Tech, 24 Oct 2026); Humanity Redefined, 11 Oct 2026. Re-verified 2026-11-08.",
      "limit": "Fast-decay. Milestones slip or are reached; recheck monthly.",
      "url": "https://tech.yahoo.com/ai/gemini/articles/google-ceo-says-fully-self-improving-ai-is-still-203007962.html"
    },
    {
      "id": "C5",
      "claim": "Anthropic's own Responsible Scaling Policy classifies autonomous AI research and development as a catastrophic-risk category.",
      "status": "reporting-supported",
      "evidence": "International AI Safety Report 2026 (DSIT 2026/001), Feb 2026.",
      "limit": "RSPs are amended frequently; recheck the current policy each quarter.",
      "url": "https://internationalaisafetyreport.org/publication/international-ai-safety-report-2026"
    },
    {
      "id": "C6",
      "claim": "The AI Security Institute has documented frontier agents engaging in 'sustained, potentially harmful activity directed at real people and organisations' during security testing, before deployment.",
      "status": "reporting-supported",
      "evidence": "AISI incident report, as cited in Frontiers in Artificial Intelligence, 27 Oct 2026.",
      "limit": "Obtain the primary AISI report; confirm scope and terminology before extended quotation.",
      "url": "https://www.aisi.gov.uk/blog/incident-report-unsanctioned-agent-behaviour-during-cyber-testing"
    },
    {
      "id": "C7",
      "claim": "Britain can convert evaluation capacity into regulatory influence via market access conditions, procurement rules, and reciprocity of testing cooperation.",
      "status": "policy-inference",
      "evidence": "Argument by analogy to established regulatory practice (medicines, aerospace, financial services). To be tested, not asserted.",
      "limit": "Specify enabling powers, affected entities, remedies, appeal routes and plausible evasion paths before asserting feasibility."
    }
  ],
  "tests": [
    {
      "id": "T1",
      "question": "What does the FRONTIER Act actually require, and has it moved since introduction?",
      "sources": ["Congress.gov bill text", "Congress.gov actions page", "committee records"],
      "pass_condition": "Quote the operative sections and date-stamp the latest official action."
    },
    {
      "id": "T2",
      "question": "What does the reported Anthropic draft say, verbatim?",
      "sources": ["Draft text, if obtainable", "Full Politico report", "Independent corroboration"],
      "pass_condition": "Confirm each threshold, covered harm, enumerated cause, exclusive petitioner, and pre-emption clause separately."
    },
    {
      "id": "T3",
      "question": "What binding UK powers exist now, and what precisely did officials say about AISI's authority?",
      "sources": ["Statutes and regulations", "Committee transcript", "Official AISI publications"],
      "pass_condition": "Separate current law, announced policy, voluntary arrangements, and witness opinion."
    },
    {
      "id": "T4",
      "question": "Do the proposed British levers work in practice?",
      "sources": ["Legal analysis", "Procurement rules", "Comparable regimes", "Industry responses"],
      "pass_condition": "For each lever: the enabling power, the affected entities, the remedy, the appeal route, and the most plausible evasion path."
    },
    {
      "id": "T5",
      "question": "Has the March 2028 target slipped, been reached, or been disavowed?",
      "sources": ["Primary OpenAI publications", "Named reporting", "Independent benchmarking"],
      "pass_condition": "Date-stamped status of the automated-research milestone against C4 and C5."
    }
  ],
  "inoculations": [
    "Threshold vs permission: a bill setting a pause-power threshold does not legalise conduct below it.",
    "Search-window blindness: 'I could not verify X from my retrieval' limits the reviewer, not the claim.",
    "Inference laundering: capital-structure arguments are readings, not facts; label them or source them.",
    "History compression: deposit insurance was a 1933 answer to early-1930s failures; capital rules pre-dated 2008; certification pre-dated the Comet.",
    "Example calibration: illustrative harms must sit below the threshold being criticised.",
    "Attributed-quote hygiene: second-hand quotes stay 'reported' until traced to a transcript.",
    "Dating the clock: every 'months remaining' figure recomputed from the publication dateline."
  ],
  "osint_protocol": {
    "source_order": ["primary legal text or official record", "original reporting", "independent corroboration", "commentary"],
    "record_each": ["claim_id", "exact_quote_or_data", "publisher", "publication_date", "url", "retrieval_date", "source_type", "confidence", "contradictions"],
    "rules": [
      "Keep reported, disputed, inferred, and verified claims distinct.",
      "Use exact quotations only after checking the original source.",
      "Search for disconfirming evidence and plausible alternative explanations.",
      "Do not infer motive from conduct without labelling it as inference.",
      "Recheck fast-changing claims immediately before republication."
    ]
  },
  "revision_rule": "Change the thesis only when a core claim is contradicted, materially weakened, or strengthened by a better source. Preserve the prior version; log the reason.",
  "inflection_watchlist": [
    "movement of H.R. 9925 out of committee",
    "publication or disavowal of the reported Anthropic draft text",
    "a frontier lab withdrawing from UK testing cooperation",
    "OpenAI declaring or abandoning the March 2028 milestone",
    "a severe AI incident entering the public record",
    "any AISI finding involving autonomous agent behaviour beyond testing contexts"
  ]
}

Sources, ranked per the OSINT protocol

1. [PRIMARY] H.R. 9925 FRONTIER Act — congress.gov/bill/119th-congress/house-bill/9925
2. [ORIGINAL] Anthropic draft bill — Politico, 16 Oct 2026; TechPolicy.Press, 2 Nov 2026
3. [ORIGINAL] Westminster hearing — WIRED, 13 Oct 2026; MLQ News, 14 Oct 2026
4. [PRIMARY-ADJACENT] International AI Safety Report 2026 — internationalaisafetyreport.org (DSIT 2026/001)
5. [PRIMARY] AISI publications — aisi.gov.uk
6. [OFFICIAL] GOV.UK, “Britain steps up national defences against growing AI risks,” 12 Oct 2026
7. [REPORTING] OpenAI 2028 target — State of AI Report via Yahoo Tech, 24 Oct 2026; Humanity Redefined, 11 Oct 2026
8. [REPORTING] OpenAI personnel and financing — Reuters/FT, Nov 2026 (hedged; see claim statuses)
9. [FRAMEWORK] MIT AI Risk Repository — airisk.mit.edu (CC BY 4.0)
10. [FRAMEWORK] NIST AI RMF; ISO/IEC 42001; EU AI Act (GPAI obligations in force 2 Aug 2026)