Anthropic built an AI hacker so powerful it scared itself — exposing the gap no government can close
Anthropic built an AI model so capable at hacking that it decided not to release it — a decision that reveals the growing gap between what private labs can build and what any government on earth is equipped to govern.