Anthropic IPO warns its AI could blackmail people, become sentient, end humanity

Anthropic’s draft IPO filing devotes 93 pages to risks that could harm its business, including unusually severe concerns about future AI systems. The company warns that highly autonomous models capable of advancing AI research could progress faster than Anthropic can assess or safeguard them, potentially creating catastrophic or existential risks.

The filing also considers how Anthropic’s operations might be affected if Claude were widely regarded as sentient or having moral status, prompting public pressure, regulation or new company safeguards. These are prospective scenarios, not claims that current Claude models are sentient or pose the described risks; the article presents them as two examples from a broader list of seven notable concerns.