Anthropic updated Claude’s usage policy in a way that marks a real shift: for the first time, the company spells out in plain text that “continued and unnecessary harassment, abuse, or cruel treatment of AI models” is outright forbidden. The change takes effect November 12. It’s aimed at extreme cases—especially where someone repeatedly “tortures” the model with no clear purpose. Once triggered, Claude may proactively terminate the conversation. This isn’t totally new territory for Anthropic. Back in Aug 2025, it said Claude Opus 4/4.1 might end chats only in rare situations to prevent potential harm—after multiple attempts to steer things back failed, and only when there’s no risk of immediate self/other harm. Now the framework is more explicit and less exploratory: abuse ends the dialogue. Normal criticism, dark fiction, and legitimate testing/research won’t be treated as punishable behavior. The update also tightens boundaries around deception (like fake accounts/news sites) and bans weapon-control and monitoring tooling, requiring continuous human oversight for AI-controlled devices. #Anthropic #Claude #AIethics #PolicyUpdate #ResponsibleAI #AIAbuse
Want to learn more? Visit Explore the world, stay updated on travel insights and international affairs, and discover authentic stories from real life
评论
发表评论