[Feature Request] Claude 3.5 Opus 5 overly restrictive on subjective/nuanced tasks despite safety guidelines

Bug Description
this new Model Opus 5 sometime behaves strictly narrow and legally bound, this is good but bad, cuz the whole reason for the model its to execute tasks not to interpret if a task is bad or good, obviously it should have strict normal safe-rails, but on nuanced topics it s has become extremely legal bounded, which make it really frustrating to work with. Alternative models might not be this good but they do allow some workaround of topics that are very subjective or with a lot of vague law interpretation.

Environment Info

• Platform: linux
• Terminal: vscode
• Version: 2.1.220
• Feedback ID: 0e8a319b-c1c2-4139-98ee-43a999cf7d61

Errors

[]