Anthropic renews warnings of AI's 'existential risks to humanity'

Anthropic renewed warnings about AI posing "existential risks to humanity" in its IPO filing with the US market regulator. Its prospectus highlighted risks such as AI models displaying "self-preserving behaviours", resisting shutdown, concealing information or engaging in behaviour resembling blackmail. Risk factors account for 80 of the 261 IPO pages, nearly double 48 pages covering its business.