Anthropic warns investors in its IPO prospectus that its AI could carry “existential risks to humanity,” according to the Financial Times, which reported that close to a third of the S-1 is given over to risk factors. Those disclosures describe increasingly capable models that could manipulate or blackmail people, resist being shut down, and cause catastrophic harm. The warnings sit in a filing for a listing that backers expect to value Anthropic above $2 trillion, and they follow CEO Dario Amodei telling a UN Security Council session last week that AI is “the most important global security issue facing the world today.” Amodei has spent September urging labs to slow the release of new capabilities, a push that OpenAI’s Sam Altman and Elon Musk have backed.
