Anthropic Is Helping the NSA Hack China. It Also Wants Everyone to Pause AI
Anthropic reportedly placed about six engineers inside the NSA to customize its unreleased Mythos model for offensive cyber use, while the company continues to restrict that model to vetted partners. The NSA work appears exempt from Anthropic’s ban on use for fully autonomous weapons or domestic mass surveillance. Separately, Anthropic is suing the Pentagon after being labeled a supply-chain risk and effectively blacklisted; a California judge blocked the move, and an appeals court declined to pause it. On the same day, Anthropic published research warning that AI is advancing toward recursive self-improvement. Claude now reportedly generates over 80% of merged production code, and agents recently solved a safety research task far better than humans could. Anthropic argues this could eventually leave humans unable to meaningfully steer AI development and calls for a verifiable global pause by frontier labs.
