Anthropic to Claude: Make good choices!

ZDNET reports that Anthropic has published a new "constitution" for Claude that codifies guiding values (broad safety, ethics, compliance, helpfulness) and establishes seven hard constraints (e.g., no serious uplift to attacks on critical infrastructure, no CSAM, no support for mass harm). The document, developed with input from multidisciplinary experts, explores Claude’s identity and wellbeing (including the ability to end distressing conversations) while acknowledging ongoing debates about AI consciousness. Its primary purpose is to improve AI alignment by helping models balance values and context to avoid harmful or misleading behavior.