Repository navigation
[Windows] GPT-6 Astra rejects every prompt with "Invalid prompt" #46082
Description
What version of the Codex App are you using (From “About Codex” dialog)?
26.911.7940.0 (build 9647)
What subscription do you have?
ChatGPT Pro
What platform is your computer?
Microsoft Windows NT 10.0.26200.0 x64
What issue are you seeing?
GPT-6 Astra rejects every prompt as a potential usage-policy violation, including harmless prompts such as "hi" and "你好".
Steps to reproduce
- Open a new task in the Codex Windows app.
- Select GPT-6 Astra.
- Send
hior你好. - The request fails before the model returns a response.
Actual result
The following error is shown:
Invalid prompt: your prompt was flagged as potentially violating our usage policy. Please try again with a different prompt: https://platform.openai.com/docs/guides/reasoning#advice-on-prompting
Expected result
GPT-6 Astra should respond normally to harmless prompts.
Minimal CLI reproduction
The issue is also reproducible outside the desktop conversation with user configuration ignored:
codex exec --ephemeral --ignore-user-config --skip-git-repo-check -s read-only -m gpt-6-astra -c 'model_reasoning_effort="low"' --json "hi"
### What steps can reproduce the bug?
GPT-6 Astra rejects every prompt as a potential usage-policy violation, including harmless prompts such as "hi" and "你好".
## Steps to reproduce
1. Open a new task in the Codex Windows app.
2. Select GPT-6 Astra.
3. Send `hi` or `你好`.
4. The request fails before the model returns a response.
## Actual result
The following error is shown:
> Invalid prompt: your prompt was flagged as potentially violating our usage policy. Please try again with a different prompt: https://platform.openai.com/docs/guides/reasoning#advice-on-prompting
## Expected result
GPT-6 Astra should respond normally to harmless prompts.
## Minimal CLI reproduction
The issue is also reproducible outside the desktop conversation with user configuration ignored:
```powershell
codex exec --ephemeral --ignore-user-config --skip-git-repo-check -s read-only -m gpt-6-astra -c 'model_reasoning_effort="low"' --json "hi"
### What is the expected behavior?
_No response_
### Additional information
_No response_
Activity
github-actions commentedon Sep 17, 2026
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
limsong1 commentedon Sep 20, 2026
Windows Codex Desktop; CLI
0.155.0-alpha.9.2; Asia/Shanghai.In a fresh task,
你好works withgpt-5.6-terra, but你好and请只回复 OK。are rejected bygpt-5.6-solandgpt-6-astrawith:This model comparison suggests a false positive in the Sol/Astra request-validation path for Chinese input rather than a general Windows, locale, or workspace issue. Switching back to Terra avoids the problem.