Investor Bill Ackman: Anthropic's Prospectus Should Address Extinction Risks
Hedge fund manager Bill Ackman stated that he looks forward to reading the risk factors in Anthropic's prospectus and questioned why the first line should not state: "Company executives believe the probability of artificial intelligence killing all of humanity exceeds 10%, revenues could go to zero, and the stock could be worthless." Evan Hubinger, head of alignment science, recently publicly agreed with former researcher Jacob Coxon's assessment, stating they seriously believe artificial intelligence could kill all of humanity, with his personal estimate of the probability being greater than 10% within the next decade; he noted that the risks of existing deployed models are low, but the concern is about superintelligence arising from recursive self-improvement, and there is currently no solution for alignment, nor is it clear that they are on the right track. CEO Dario Amodei previously placed the probability of a "very, very bad" outcome at 10% to 25%. Coxon stated he left after pre-training at OpenAI and Anthropic, indicating both are rushing towards self-improving superintelligence. Another head of oversight mentioned that the more senior the employee, the more concerned they are.
Anthropic is preparing for an IPO, with reports indicating that a confidential draft has been submitted, and the documents may be made public in the coming weeks, or trading may be delayed until November to include third-quarter figures. Annualized revenue estimates have risen from about $65 billion in July to potentially over $100 billion by year-end, with discussions of a potential valuation around $2 trillion. The company is a public benefit corporation, and whether to retain long-term interest trusts and dual-class shares will affect shareholder rights. Amodei also issued a statement calling for a slowdown in frontier capabilities to allow safety verification to catch up. Foreign safety agencies, including those in the UK, reportedly have not received complete shares of the latest models.
Securities law requires the disclosure of known significant risks. The Financial Times has discussed how to incorporate civilization-level disasters into the prospectus and has had Claude draft example paragraphs. Extinction risks are typically not insurable. Ackman framing revenue going to zero as a risk factor translates researchers' probability assessments directly into securities text.
In market mechanisms, buyers are institutional investors who need to price valuations before the IPO, while sellers are labs that are selling rapidly growing revenues while publicly disclosing extinction probabilities. Funds are flowing from the primary market into unpriced liability gaps. The event is driven by Ackman's posts combined with IPO expectations. Beneficiaries are underwriters and issuers who can write risk factors into a sellable narrative; those under pressure are the legal teams that must address executives' stated probabilities in the prospectus, and the pricing models that must place a 10% species risk alongside a trillion-dollar valuation in the same document.
The Trump administration downplayed safety risks at the same time and discussed appointing an AI czar, which is inconsistent with internal lab assessments.
Source: Public Information
ABAB AI Insight
Ackman's pressure is not about philosophical debate, but about the first line of the S-1. Once the 10% to 25% probability comes from the CEO and the head of alignment, the securities law's "known trends and uncertainties" can hardly be framed as standard competitive risks. The public benefit corporation and trust were originally meant to prioritize mission over shareholders, but going public translates the same mission into tradable equity, making risk factors the seam between the two. With revenues soaring towards $100 billion, if the extinction paragraph is written lightly, future lawsuits will cite Hubinger's original words; if written heavily, the pricing model must incorporate species risk into the discount rate.
The capital path is for underwriters to sell a $2 trillion discussion value with a growth story, while the safety team maintains regulatory dialogue with the same mouth. The open letter to slow down the frontier and the rush to IPO run parallel, indicating that safety claims are being used in policy arenas, while timelines are being used in capital markets. Not providing the latest models to foreign assessment agencies effectively splits information disclosure into internal reports and external prospectuses.
Benchmarking against product liability clauses in tobacco and opioids, Boeing's risk updates after safety incidents, and the unprecedented "product or eliminate buyers" phrasing in software prospectuses, frontier labs are in the pre-IPO stage of productizing doomsday probabilities, with neither bonds nor equity prepared for uninsurable tail risks.
Structural changes belong to the transfer of pricing power. Pricing power temporarily shifts from revenue multiples to risk factor wording: whoever can write 10% in a way that does not affect issuance and can withstand post-facto citation temporarily controls valuation. The mechanism is that once researchers vocalize the probability, lawyers must choose either to include it in the document or explain in court why it was not included.
ABAB News · Cognitive Law
- The extinction probability given by executives will become a pricing variable in the prospectus.
- Revenues in the hundreds of billions and species risks written in the same document, with no ready formula for the discount rate.
- Safety statements sold to regulators, growth stories sold to the IPO window, meeting in the seam of risk factors.