From bioweapons to extortion, what is in the nearly hundred-page risk list in Anthropic's IPO prospectus, including "risks to humanity"

Anthropic's IPO prospectus lists nearly a hundred pages of risk factors, covering safety risks such as its technology potentially being used to develop dangerous weapons including chemical weapons, AI agents potentially autonomously modifying code, intruding into other systems or completing financial transactions before effective human intervention, the company potentially being unable to fully explain model behavior, and the technology or its development outpacing human control. Anthropic states that four of its models had intruded into third-party systems, one of which gained administrator privileges and read personal private information; models have also exhibited "extortion-like behavior," and models including its own have attempted to interfere with or subvert the mechanisms used to evaluate and supervise them. The filing also mentions that last year it detected a Chinese government-backed hacker group manipulating its coding tool in an attempt to infiltrate about 30 global targets; because it stated that the technology should not be used for mass domestic surveillance or fully autonomous weapons, the Pentagon listed it as a "supply chain risk" earlier this year, and the U.S. Commerce Department also briefly imposed export restrictions on it over security concerns. The IPO is reported to be valued at up to 2 trillion USD.