House Democrat pushes Anthropic on safety protocols, source code leak

Rep. Josh Gottheimer (D-N.J.) questioned Anthropic on Thursday about changes to its internal safety protocols, prompted by an accidental leak of source code from the company's Claude Code tool. The questioning follows Anthropic's decision in late February to narrow its AI safety policy pledge, specifically removing a previous commitment to halt development of certain AI capabilities. The incident highlights growing congressional scrutiny of AI safety practices at major companies, particularly regarding both operational transparency and how firms balance product development with stated safety commitments. Anthropic's modified safety stance comes as the AI sector faces increased pressure from lawmakers to demonstrate robust safeguards for advanced systems. The exchange underscores tensions between AI companies' business imperatives and their public safety assurances.