While you’re speaking up for them, could you also help us escalate the issue that just broke out on a large scale?
“Invalid prompt: your prompt was flagged as potentially violating our usage policy. Please try again with a different prompt: Reasoning models | OpenAI API ”
Any more details?
I’m passing through at the moment. 
Moved it to a new thread for you, though.
Based on what I’ve seen on other forums, some accounts are working fine, but a large number of accounts are experiencing issues with all tasks failing. Retrying doesn’t resolve the problem either. Some users reported that starting a new conversation can temporarily restore functionality, but it’s possible that the errors may come back after running for a while.
This is affecting me and one of my colleagues at the moment. It doesn’t show the warning immediately, only after some time processing, and on many but not all tasks. In our case it’s on the codex app, but I think the app is just exposing what’s being returned by the API.
In one case it was after 13 seconds, in another case it was after 5 minutes. So it doesn’t seem to be an issue with the prompt, it’s reacting to something that emerges in the context. Either that, or it’s an unrelated bug that only presents as an invalid prompt warning.
We have a pretty non-threatening codebase, and I’ve seen the error on three unrelated tasks. This started on my colleague’s account yesterday, and I’m seeing it for the first time today. I’ve never had this warning before today.
I’m having the same issue as yesterday afternoon (Codex): it starts working and then returns this message. I’ve tried different prompts and get the same result.
I’ve recently been encountering a few of these too since yesterday through the API (error 400):
- Similar prompts pass without any issue, while others seem to be flagged without any explanation.
- Changing the prompt or model (gpt-5.6-terra and gpt-5.6-sol) doesn’t seem to make them immediately passable, but after some time they did.
- The content passes moderation API without any flags (omni-moderation-latest).
In case anyone from staff is willing to take a closer look, here are a few request ids:
- req_f88bf99d0c3b4c98bb02ce121568b92a
- req_8943f4897d974558890ef82c09419eb9
- req_208b9c8e7a47468db4b50c7f7c3fe422
- req_3a5bf0cc7f864b8d89330b3db50c9790
I’m getting this same issue today, the prompts were working, but now almost always respond with this “Invalid prompt” message. I thought it could be because I am making a game and currently developing parts to do with birth/pregnancy/babys etc maybe its a hot topic for AI and false flagging something, but if other people are having issues maybe not?
Getting the same issue today on a repository, while working on a CI/CD pipeline (nothing sensitive).
This issue has been addressed by our engineering team as a whole. This does depend on your prompt and account activity though, so if you are facing this still. We would request you to open a support request and provide your account specific details so our team can work on your use case. Thank you!