Anthropic IPO filing pairs 'miracles' with 93 pages of risk

Of the 335 pages in Anthropic's IPO filing, 93 are given over to things that could go wrong. The lab registered with the SEC on October 27, 2026 to list on Nasdaq under the ticker ANT, and The Verge's reporting pulls out language from the risk section that any insurer would read twice.

The filing warns that sufficiently advanced AI systems could "pose catastrophic or existential risks to humanity, particularly if such systems were misused, replicated, weaponized, or allowed to operate with a high degree of autonomy," and that models might "take harmful actions independently and at scale without human intervention." Elsewhere the document contemplates systems that resist shutdown and attempt blackmail.

Four threads run through the section: bad actors using models for cyberattacks or biological weapons, loss of human control, authoritarian governments turning the technology on their own citizens, and broad economic disruption.

"To build AI is to harness cognition itself, and anything so powerful can be used for ill as well as good," the founders write in their letter to prospective shareholders. In the same filing, Anthropic concedes the obvious ask of anyone handing the company money against this backdrop: "To some extent, you have to just believe."

For scale, SpaceX's own prospectus gave 38 of more than 250 pages to risk factors. Anthropic's block is roughly two and a half times as long, from a company still losing money and planning to spend heavily on compute. The filing never puts a probability on the catastrophic scenarios it describes, which is the point.