Filed for an IPO, disclosing risks that include potential catastrophic harm from AI. (Oct. 27)

Anthropic filed its IPO prospectus on Oct. 26, warning that advanced AI could pose “catastrophic or existential risks to humanity” if deployed without safeguards. The 335-page filing devotes 93 pages to risks, including models acting autonomously, resisting shutdown, manipulating people or producing behavior resembling blackmail; it also cites misuse for cyberattacks and biological weapons. Anthropic cautioned that insurance might not cover losses from a catastrophic AI incident. The disclosure makes safety a material investor concern as well as a product issue: an incident could bring lawsuits, regulatory action and reputational damage across the sector, affecting customers that rely on Claude. The prospectus also reveals the scale of the business behind the warning: first-half 2026 revenue reached $16.3 billion, up about 1,300% year over year, while net losses were $80.7 billion, largely reflecting changes in the value of financial liabilities. Anthropic reported more than 14,000 enterprise customers.