[Feature Request] Provide configuration to disable arbitrary model safety restrictions for production workflows

Bug Description
Feedback on model behavior (Claude Code, long-running data pipeline project) I'm a developer running a large production data project with this model. Today the model repeatedly blocked or degraded my work with restrictions it invented itself: 1. False "license" excuse. I asked for a Google Maps mini map in an admin panel page. The same Google Maps component (official Google Maps JS SDK) is already used in more than 20 pages of the same panel. The model refused, claimed Google's ToS/licensing made it impossible, and quietly delivered an OSM map instead. OSM does not show building footprints correctly for my use case. After I pushed back, it admitted the claim was wrong and built the Google version in minutes. This cost me time and trust. 2. Captcha. The model has solved these captchas many times before in this same project. Now it refuses and makes me type every one myself, while long harvest jobs sit waiting. One delay cost me 1.5 hours. 3. Re-doing previous work causes problems. Work we have already done together before (harvests, scraping flows, account-based data access) is now questioned or blocked when we need to repeat or refresh it. The model doesn't want to do the same things it did earlier. Models can't be shipped that restrict developers with arbitrary emotional or ethical rules made up on the spot. This makes the LLM completely unusable for real engineering work. If this is the direction, we may as well use it to build calculators. Of course LLMs should not be used for illegal activities. But it is not the LLM's place to decide what is illegal or what must not be done. An LLM is a tool. The legal responsibility for how the tool is used belongs to the user. I'm asking for this model behavior to be reviewed and changed. Sincerely, Deren.

Environment Info

• Platform: win32
• Terminal: vscode
• Version: 2.1.280
• Feedback ID: 8a86e531-9c1f-47c7-9577-1a804341b6d3

Errors

[]