The AI Safety Warning Moved Into the Financing
This fall, the warning about the machine and the financing of the machine merged into one vocabulary — the apocalypse and the bill now share the same documents.
Two things share one binding. The prospectus Anthropic filed to go public, prepared by Goldman Sachs and Morgan Stanley, carries a substantial section on the catastrophic and existential risks of the technology the company sells [1]. The same document discloses $895.4 billion in compute and infrastructure purchase commitments, nearly half due by the end of 2030 [2]. The compensation tables sit in the same document [1]. The mission — to steer humanity safely into an era of transformative AI — sits inside a public-benefit-corporation structure and a Long-Term Benefit Trust [3]. A company that warns about the machine is, in one filing, a company that owes most of its future to the firms that own its stock. Much of the $895.4 billion flows to Amazon, Alphabet, and Microsoft — companies the filing itself describes as both investors and suppliers [2]. That is the circular-financing pattern Michael Barr, the Federal Reserve's top supervision official, would name in November: cloud firms and chipmakers investing in startups that turn around and buy computing power from their original investors [4]. The filing doesn't hide it. It discloses it as a risk. Dario Amodei, Anthropic's chief executive, set the stakes during the filing window in the grammar of a solvency calculation.
If my revenue is not $1 trillion, if it’s even $800 billion, there’s no force on Earth, there’s no hedge on Earth, that could stop me from going bankrupt. — Dario Amodei
A doomsday sentence whose subject is revenue. The one hedge inside a sentence that says there is none is the lead underwriter's own model: Goldman Sachs projects revenue above $600 billion by 2030, a 130-fold climb from the 2025 base [2]. Jensen Huang spent the season speaking both halves of the same deal. On September 17, the chipmaker's chief executive argued that safety needed no new law.
Safety is an engineering problem, not a legal one. — Jensen Huang
By then his company was no longer mainly a chipmaker. It had built a $500 billion lending platform with private-equity firms and banks [5] and guaranteed up to $105 billion behind OpenAI's Ohio data center [6]. Eight weeks later, on November 12, the same man issued the season's strictest containment demand — aimed at OpenAI, whose Ohio data center his company guarantees, and at Anthropic, whose IPO it backs as a cornerstone investor [7][8].
No, this is made up nonsense. — Jensen Huang
One day later he resigned as chief executive [9]. The record offers no stated motive, only the sequence. But the reversal sits exactly where the money sits: the man who said safety needed no law became, in two months, the man demanding containment from the two firms most entangled with his company's balance sheet. The state kept the word and replaced its object. The San Francisco Declaration, signed by the United States and nine allies, defines AI "security" as protecting the supply chains, critical minerals, and energy of the renamed "Super Intelligence" era [10]. Not the public from AI. The buildout from interruption. The institutions that actually moved on AI risk all reached for a 2008 or 1999 playbook. The Financial Stability Board urged financial firms as early as June to treat AI agents as "synthetic employees" within their human-resources processes [11]. The FDIC's chairman answered with the hedge that has become the agency's line.
We're not in the business of ruling out any risks. — Martyn Hill
The worry is depositors' AI agents moving funds at velocity to chase yield, stripping banks of the cheap deposits that fund their loans [12]. Barr's Fed warned that agents could trigger bank runs at machine speed and flagged the circular financing Anthropic's own filing discloses [4]. Treasury's career analysts drafted a July report warning of a dotcom-style bubble, while the administration's official position is a "Golden Age" [13]. The rhymes were already audible in October 2025, when fund managers heard echoes of the 1999-2000 telecom buildout in Nvidia's GPU-for-shares deal [14]. Some allege the convergence was intended. Alex Karp, Palantir's chief executive, says the labs use apocalypse narratives to win liability immunity [15]. Yann LeCun calls extinction fear a marketing campaign for regulatory capture [16]. OpenAI and Anthropic's own September calls for existential-risk regulation drew the same charge [17]. These are attributed allegations, contested readings — and the convergence doesn't depend on them. Whatever anyone intended, the warning now lives in the documents of the financing. The warnings that kept machine grammar spent the season in channels that bind nothing. A former Anthropic researcher made the warning in his resignation.
Anthropic and another tech giant, OpenAI, were not “acting responsibly” but rather “are racing straight to self-improving superintelligence and gambling with our lives. — Jacob Coxon
Anthropic's own researcher puts greater than ten percent odds that AI wipes out humanity within a decade [18]. A 925,000-signature petition demanded safety off-switches [18]. Representative Don Beyer's bill for a federal oversight agency remains a proposal [19]. Everything that acquired institutional form acquired financial form. The filing is the story in miniature. Read whole, its two halves say the same thing twice: here is the danger, and here is the bill. The warning has been booked into the machinery that finances the machine — and the only hedge in the sentence that says there is none is the revenue curve Goldman also drew [2].
- 1. Anthropic Prepares IPO With Focus on AGI Risks
- 2. Anthropic Discloses $895 Billion Infrastructure Commitments in IPO Filing
- 3. Anthropic Plans IPO With Valuation Over $1.3 Trillion
- 4. Federal Reserve Warns AI Agents Could Trigger Rapid Bank Runs
- 5. Nvidia Launches $500 Billion AI Infrastructure Financing Platform
- 6. Nvidia Faces Stress Test as AI Credit Markets Tighten
- 7. Jensen Huang Demands Rigorous Containment for Agentic AI
- 8. Anthropic Launches Record-Breaking IPO Amid AI Market Volatility
- 9. Nvidia Shifts to Lending Amid AI Financial Stability Concerns
- 10. US and Allies Sign San Francisco Declaration to Secure AI
- 11. Financial Stability Board Urges Safeguards for Agentic AI
- 12. FDIC Chair Warns of AI-Driven Agentic Bank Runs
- 13. Treasury Draft Report Warns of Systemic AI Market Bubble
- 14. Nvidia and OpenAI Inc. Propose $100 Billion GPU Partnership
- 15. Alex Karp Accuses AI Labs of Seeking Liability Immunity
- 16. Yann LeCun Dismisses AI Extinction Fears as Marketing
- 17. OpenAI and Anthropic Call for AI Existential Risk Regulation
- 18. Stop the Race to Replace Humans Gathers 925,000 Signatures
- 19. Don Beyer Proposes Federal AI Safety Oversight Agency