I propose that GPT models be able to explicitly explain when a user’s request violates OpenAI’s use policies.
Specifically, the model should:
- state that the request violates policy;
- cite the relevant section (e.g. “Use Policy section X.Y”);
- briefly explain the nature of the violation;
- and, optionally, suggest how to rephrase the request in a compliant way.
Benefits:
- clearer communication with users;
- less frustration and confusion;
- promotes responsible use and learning;
- increases trust in AI transparency.
Context: this is especially important for users who intentionally ask for policy clarification and want to understand the limits of the system rather than exploit it.
P.S.: Credit to 4o model for better formulating purposal