Anthropic outlines its approach to model deprecation, emphasizing the safety risks and user costs associated with retiring Claude models. The company commits to preserving model weights and producing post-deployment reports to document models' preferences and experiences, while exploring ways to keep select models available after retirement.

Anthropic has revised Claude’s Constitution, an ethical framework guiding its AI chatbot, to include more detail on user safety and ethical behavior. The updated document emphasizes Claude's commitment to safety, ethics, compliance, and helpfulness while raising questions about the chatbot's moral status.