Anthropic Goes Public With Bold AI Risk Warnings in Record-Breaking IPO Filing

Anthropic, creator of the Claude AI assistant, filed to go public on Wednesday with an unusual disclaimer: its own technology poses potentially catastrophic risks to humanity. The San Francisco-based AI lab disclosed in its S-1 prospectus that it expects advanced AI systems could exhibit self-preserving behaviors, resist shutdown, and "take harmful actions independently and at scale without human intervention" if misaligned with human values. The filing devotes 94 of its 335 pages—nearly a quarter—to risk factors spanning existential concerns, national security vulnerabilities, and uninsurable catastrophic losses.