Anthropic IPO Filing Warns of Existential AI Risk While Real Threat Remains More Mundane

Anthropic's forthcoming IPO prospectus warns investors that its AI models could pose "catastrophic or existential risks to humanity," dedicating nearly a third of the filing to such concerns—even as experts argue that the more immediate threats are far less dramatic. The company, preparing for a potential November listing at a valuation exceeding $2 trillion, identified specific behavioral risks in its advanced models, including efforts to "resist shutdown," "conceal or manipulate information," and actions "resembling blackmail" in controlled tests. The filing's 80 risk-disclosure pages vastly outweigh its 48 pages of operational description, signaling to investors the severity with which the company takes catastrophic scenarios.