Anthropic says some Claude models can now end ‘harmful or abusive’ conversations
-
最新报道：Anthropic has introduced a new feature allowing its Claude Opus 4 and 4.1 models to terminate conversations in extreme cases of harmful interactions, such as requests involving child exploitation or terrorism. The company emphasizes this measure aims to protect the AI's "model welfare" rather than human users, citing observed "distress patterns" during testing. Claude will only end chats after repeated redirection attempts fail or if explicitly asked, excluding cases where users may harm themselves or others. Terminated conversations still permit new chats or edited branches. Anthropic calls this an experimental approach to mitigating potential risks to AI welfare.
下一篇
AI-powered stuffed animals are coming for your kids
- AI-powered plush toys like Curio's Grem and Grok are being marketed as screen-free alterna...
24小时热点
热点专题
407834