Anthropic has cautioned potential investors that highly advanced artificial intelligence could present "catastrophic or existential risks to humanity", according to reporting by the Reuters news agency and the Financial Times.
The reported language appears in an IPO prospectus that has not yet been made public. The filing is said to acknowledge the possibility that future AI systems might display "self-preserving behaviours"—a phrase indicating concerns about systems acting to protect their own continued operation or goals.
The disclosure comes as Anthropic prepares for a potential initial public offering that market reports have valued at up to $2 trillion (around £1.5 trillion). Details of the proposed flotation, including timing and the full contents of the prospectus, have not been released publicly.
The warning echoes earlier calls from Anthropic for a slowdown in the rapid development of AI capabilities, a stance that has also been voiced by several other firms and researchers in the sector. The reported admission in the IPO filing adds that risk assessment directly into the company’s communications with prospective investors.
Reuters and the Financial Times were the first to report the details of the filing. Until the prospectus is formally published, independent verification of the full text and any additional context in the document is limited.