closing the loop

Anthropic’s public S-1, filed on 26 October 2026, devotes 94 pages to risk factors that read like an AI doomsday primer: AI that helps build its successors, autonomous agents, self-preserving behaviour, weapons uplift, cyberattacks and robots. We quote each mechanism with page numbers, show what the filing says Anthropic has already done, and explain what it means for businesses using Claude.