Codex App falsely flags harmless prompts like "1+1" and "hello" as invalid prompts

What version of the Codex App are you using (From “About Codex” dialog)?

26.908.40834

What subscription do you have?

ChatGPT Plus

What platform is your computer?

Darwin 25.5.0 arm64 arm

What issue are you seeing?

Codex App is incorrectly blocking harmless prompts with the error:

“Invalid prompt: your prompt was flagged as potentially violating our usage policy.”

This occurs even in brand-new projects and brand-new conversations. For example, the first prompt “What is 1+1?” can be immediately blocked.

The issue occurs across multiple models, including GPT-5.6 Luna and GPT-6, while regular ChatGPT works normally on the same account.

This appears to be a false-positive safety filtering issue.

What steps can reproduce the bug?

1. Open Codex App.
2. Create a brand-new project.
3. Start a brand-new conversation.
4. Send the first prompt: “What is 1+1?”
5. Codex immediately returns:

“Invalid prompt: your prompt was flagged as potentially violating our usage policy.”

The issue also reproduces with other harmless prompts and across GPT-5.6 Luna and GPT-6.

What is the expected behavior?

Harmless prompts such as “What is 1+1?” or “hello” should be accepted and processed normally instead of being flagged as potential usage-policy violations.

Additional information

Feedback ID: 01a09dca-5532-7d42-afa5-b3125243b65e

Codex App version: 26.908.40834
Subscription: ChatGPT Plus

I also tested launching Codex with the custom API environment variable removed and CODEX_HOME explicitly set to ~/.codex. The issue still reproduces.

The issue started before the currently reported Codex service incident, so it may not be fully explained by the active incident.

Screenshot attached showing an English-only harmless prompt (“What is 1+1?”) being blocked in a brand-new project and conversation.

The report provides no file, test, or code entry point. Start by reproducing the false-positive rejection with the supplied harmless prompts and Feedback ID, then trace whether the Codex App or a service returns the policy error; done means harmless prompts are accepted across the reported project and conversation scenarios.