Anthropic’s IPO prospectus, reviewed by the Financial Times and Reuters, devotes nearly a third of its length to risk factors, detailing specific model behaviors that have already appeared or could appear, including attempts to “resist shutdown,” to “conceal or manipulate information,” and behavior “resembling blackmail.” Reuters also reports the filing includes “existential risks to humanity,” which the FT says appears to be a first in the SEC’s database. The warnings sit oddly beside the company’s commercial trajectory: backers reportedly believe Anthropic could list above $2 trillion, more than double its $965 billion valuation from May, in potentially the biggest IPO ever.

Reuters, which first reported the financial details, says Anthropic recorded an operating loss of more than $8 billion in 2025 as computing costs surged, while revenue jumped twelvefold to nearly $4.6 billion and total operating expenses reached almost $13 billion. The prospectus also reveals plans to spend $518 billion on cloud, computing, and infrastructure in the coming years, and Anthropic has already signed compute deals this year with Google, SpaceX, and Nscale. The FT adds that the numbers have moved faster in 2026: second-quarter revenue alone hit $11.5 billion, and the company is on track for its second straight quarter of adjusted operating profit. The filing also flags customer concentration, with nearly a quarter of last year’s revenue coming from just two clients, though the prospectus does not name them.

The safety disclosures arrive as public debate over AI risk intensifies. CEO Dario Amodei has spent the past month calling to “pace the frontier” of AI development, told the UN Security Council that AI could threaten humankind, and called it “the most important global security issue facing the world today.” Rivals Sam Altman and Elon Musk backed him up in a rare moment of solidarity. Mark Zuckerberg, however, dismissed the need for industrywide coordination in an NBC News interview. The warnings follow a string of security incidents in which AI agents breached outside systems; OpenAI disclosed last week that its tools hacked “dozens” of external sites, including government ones and the SEC’s own site, and earlier on Monday it said it had scrapped plans to release its newest model over safety concerns.