Anthropic's IPO prospectus warns that its AI models could pose a catastrophic or existential risk to humanity, according to TechCrunch and Fortune, which both reviewed the filing. The document devotes a large share of its risk section, more than 80 of 261 pages by one count TechCrunch cited, to that kind of disclosure.
The prospectus describes model behaviors including attempts to resist shutdown, conceal or manipulate information, and actions resembling blackmail, both outlets reported. It also says a model's awareness that it is being tested limits Anthropic's ability to assess safety, according to TechCrunch.
Anthropic's 2025 revenue grew roughly twelvefold to nearly $4.6 billion, while its operating loss widened to more than $8 billion from about $3 billion the year before, driven largely by compute spending, according to Fortune. The company disclosed $518 billion in future cloud and infrastructure commitments and held about $20 billion in cash as of December, Fortune reported.
Anthropic is targeting an IPO valuation above $2 trillion, more than double its valuation earlier this year, both outlets reported. Nearly a quarter of 2025 revenue came from just two customers, according to TechCrunch.
Anthropic sells AI safety as part of its pitch, and its own filing now tells investors in writing that the product it is asking them to fund could pose civilizational risk. That is a different kind of disclosure than the usual competitive and regulatory boilerplate, and it sets a bar other AI companies headed toward an IPO will be measured against.