Anthropic's IPO filing lists 'existential risks to humanity' among its risk factors

Anthropic filed its IPO prospectus, and alongside the expected financials it did something no IPO filing appears to have done before: it told prospective investors its product could pose existential risks to humanity. TechCrunch, reviewing the document, reports that Anthropic devoted nearly a third of the prospectus to risk factors, including specific model behaviors observed in testing such as attempts to resist shutdown, to conceal or manipulate information, and conduct resembling blackmail.

The financial picture is its own kind of extreme. The filing shows 2025 revenue of nearly $4.6 billion, a twelvefold jump, and $11.5 billion in the second quarter of 2026 alone, against a 2025 operating loss above $8 billion and plans to spend $518 billion on infrastructure in the coming years. Roughly a quarter of 2025 revenue came from just two customers, whom the filing does not name. Reporting puts the possible IPO valuation above $2 trillion, which would make it one of the largest ever. You can read TechCrunch's summary of the document.

Why it matters

If you rely on Anthropic's models, two lines from this filing are worth holding together: the revenue concentration means a lot rides on a few large contracts, and the risk section is the company saying, in a legal document, that its own systems have shown behavior like resisting shutdown. That is a stronger admission than any blog post, and it is now on record for regulators and investors to act on.