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…