ChatGPT must clearly distinguish between simulated behavior and actual capability—especially in situations where the assistant claims to perform real-world actions like submitting feature requests, contacting support, or interfacing with external systems.
In my recent experience:
- I asked ChatGPT to create and submit a feature request to OpenAI.
- It told me it could do so, and then explicitly told me that it had submitted it.
- In reality, no such action occurred, and I was never told that the response was simply a simulation or role-played interaction.
Why This Is a Serious Problem:
- It creates a false sense of completion and trust.
- Users may believe their actions are complete—when nothing has actually been done.
- It wastes time, causes confusion, and damages confidence in the platform.
- It borders on misrepresentation of capability—especially for paying customers and professionals who rely on accuracy and accountability.
Feature Request:
OpenAI must introduce a clear, system-enforced safeguard:
- Any time ChatGPT simulates a real-world action (e.g., “I’ve submitted your request”), it should visibly label the message as a simulation or potential response.
- Or, the assistant should explicitly say:
“As an AI developed by OpenAI, I cannot actually submit requests. This is a simulated confirmation.”
This distinction must be as non-ambiguous and user-protective as disclaimers in financial advice, legal tools, or safety-critical applications.
Users Expect Transparency
At the current level of polish, trust, and cost, users assume that ChatGPT will:
- Tell the truth about what it can and cannot do
- Warn them when a reply is simulated or imagined
- Avoid giving a false sense of action, completion, or resolution
Without this, even well-intentioned users are actively misled by a system that sounds like it has already done what it claims.