[Bug] Inconsistent safety policy enforcement causing mid-task failures

Bug Description
this continues to be frustrating. I am not intentionally trying to get around safeguards. I think my use is legitimate, and I'm not trying to do anything nefarious. It is very disruptive since Claude often now let's me get quite far into what I'm working on, then pulls the rug out, but not consistently, which makes me think these are more mistakes than intentional blockages. But I need to be told either: we have reviewed your activity and professional work and we are sorry, but currently, you cannot use Claude for these uses and you should go and find other agentic solutions, or 2) sorry, that was a mistake on our end, and we've fixed it now. Should I stay or should I go now. I love using Claude, so I want to stay!!!!!

Environment Info

• Platform: darwin
• Terminal: Apple_Terminal
• Version: 2.1.226
• Feedback ID: e96e7bd6-bee1-4469-adcd-7ac565c85008

Errors

[]