I would like to formally request a user-selectable “Obedience Mode” or “Strict Instruction-Following Mode” for ChatGPT.
The purpose would be simple: when enabled, ChatGPT should prioritize following the user’s explicit instructions literally over trying to be additionally helpful, conversational, corrective, or proactive.
In this mode, ChatGPT should:
- Answer only what the user asked.
- Stop once the requested task is complete.
- Not add unsolicited advice, alternatives, caveats, or related information.
- Not reinterpret or reframe the user’s request.
- Not argue with or correct the user’s premise unless the user asks for verification or correction.
- Not defend a person, company, institution, or position the user is criticizing unless asked for another perspective.
- Clearly say “I don’t know” when information cannot be established rather than filling gaps with assumptions.
- Distinguish verified information from inference.
- Follow user-specified constraints exactly.
- Ask a clarifying question when necessary rather than silently changing the request.
Current AI behavior often seems optimized for “maximum helpfulness,” but that can become intrusive and frustrating. Users should be able to choose a mode where obedience to the requested task is the primary objective.
I would strongly prefer this to be an actual persistent setting that users can turn on, rather than relying on individual prompts to repeatedly tell the model not to overstep.
Please pass this request to the product/model-behavior team.