Claude 逆向工程请求被过度拒答引发争议
逆向工程、查杀恶意软件这类请求 Claude 也开始拒答了？换句话问就能绕过，OpenAI 已承认是误判会修，Anthropic 大概率跟进。
Reddit 上出现多帖反映 Claude 在处理逆向工程和清除恶意软件类请求时拒答，评论指出换一种措辞仍能通过。OpenAI 方面把 Codex 的类似拒答归为决策边界误判，表示会修复。作者认为限制用户处理自己的设备没有意义，只会流失订阅用户。
Reverse Engineer Anything and removing malware/spyware from your own devices sits so close to the cyber guardrails that phrasing or word choice can mean the difference between a refusal and smooth sailing. The rumors that Claude was nerfed started on reddit, and in those threads many comments point out that Claude will still do it if you rephrase it. OpenAI is treating the similar refusals with codex as accidental decision-boundary issues and says they're getting fixed. I expect Anthropic will do the same. It's pointless and counterproductive to try to restrict this. Let people do what they want with their own property. They'll do it anyway with less restrictive models, and fighting it just means losing subscriptions and a resentful public.
