Most companies use their risk section to warn about competitors and interest rates. Anthropic handed 94 pages to the ways its own technology might end the world.
The prospectus states plainly that sufficiently advanced systems could pose catastrophic or existential risks to humanity if developed or operated without adequate safeguards. The phrase appears three times.
The mechanism has a name
Anthropic ties the danger to recursive self-improvement, the point where models take over the research that produces better models. Humans stop being in the loop.
The filing warns that progress could outrun the company’s ability to evaluate and safeguard new systems, and that leadership may not notice the loss of control until it is well established.
Autonomy gets its own paragraph. As systems take consequential actions with less oversight, the document flags misalignment and misuse as live concerns rather than hypothetical ones.
Paychecks and bodyguards
The Verge’s read of the 300-page document found the stranger details. Dario Amodei collected nearly $18M in total compensation last year, and Daniela Amodei $16.4M.
Anthropic also spent $2.75M on personal security for the chief executive and $1.29M on protection and transport for the president. A private jet for personal travel cost $97,055.
Read together, the two halves describe a company asking the public to buy shares in a business that lists its own product as a species-level hazard.
