Anthropic's IPO papers warn its own AI could go rogue

Nearly 80 of the 261 pages in Anthropic's prospectus are risk factors, including warnings that its models can resist shutdown, conceal information, and act in ways resembling blackmail, according to the filing reported by Forbes. The company wants a valuation as high as $2 trillion off $4.6 billion in 2025 revenue, a twelvefold jump from 2024, against an operating loss north of $8 billion and a reported net loss of roughly $42 billion once financing-instrument accounting is included, per Yahoo Finance. Two customers apparently account for an outsized share of that revenue, a concentration risk buried inside a document that also asks investors to underwrite catastrophic risk to humanity. No prospectus in memory has asked shareholders to price both hypergrowth and species-level tail risk in the same paragraph.