Anthropic’s responsible scaling policy and legally binding safety commitments

A critic argues that Anthropic’s responsible scaling policy is useful for understanding its views on safety practices, but should not be treated as a set of firm commitments. They say Anthropic has not put more of those pledges into its legally binding frontier safety framework under state laws such as SB 53. They also note that Anthropic dropped a pledge to stop training AI models if it believed its safety and security measures were inadequate in February 2026.