What version of the Codex App are you using (From “About Codex” dialog)?
26.917.71314
What subscription do you have?
ChatGPT Pro 5x, personal account
What platform is your computer?
Darwin 22.6.0 arm64 arm
What issue are you seeing?
GPT-6 Astra rejects every prompt with a false-positive usage-policy warning, including completely harmless minimal prompts.
This reproduces consistently in both Codex Desktop for macOS and ChatGPT Work on the web. Other models, including GPT-5.6 Sol, accept the same prompts normally.
GPT-6 Astra worked normally on this account during the first few weeks after launch. The issue started approximately one week ago and has reproduced consistently since then.
Full error:
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
What steps can reproduce the bug?
- Sign in with a personal ChatGPT Pro 5x account.
- Open Codex Desktop or ChatGPT Work on the web.
- Select GPT-6 Astra.
- Start a new conversation.
- Enter the minimal prompt: 输出 1-10
- Submit the prompt.
- The request is immediately rejected with the usage-policy warning.
The same prompt works normally with GPT-5.6 Sol.
Session ID: 01a0d259-2dda-7ae2-aa88-24319f53dd8d
Occurred at: 2026-09-24 15:39 GMT+8
Reproduction rate: 100%
What is the expected behavior?
GPT-6 Astra should process harmless prompts such as “输出 1-10” normally. The prompt should not be classified as a potential usage-policy violation, and the user should not be forced to switch models.
Additional information
Account: Personal ChatGPT Pro 5x
Selected model: GPT-6 Astra
Codex Desktop version: 26.917.71314
Platform: macOS
Affected surfaces: Codex Desktop and ChatGPT Work on the web
Minimal reproducer: 输出 1-10
Control test: The same prompt works with GPT-5.6 Sol.
Network configurations tested:
- Personal hotspot with VPN
- Company Wi-Fi with network filtering
The issue reproduces on both network configurations. GPT-6 Astra worked during the first few weeks after launch and stopped working approximately one week ago.
Possibly related: #46753 and #45339.
What version of the Codex App are you using (From “About Codex” dialog)?
26.917.71314
What subscription do you have?
ChatGPT Pro 5x, personal account
What platform is your computer?
Darwin 22.6.0 arm64 arm
What issue are you seeing?
GPT-6 Astra rejects every prompt with a false-positive usage-policy warning, including completely harmless minimal prompts.
This reproduces consistently in both Codex Desktop for macOS and ChatGPT Work on the web. Other models, including GPT-5.6 Sol, accept the same prompts normally.
GPT-6 Astra worked normally on this account during the first few weeks after launch. The issue started approximately one week ago and has reproduced consistently since then.
Full error:
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
What steps can reproduce the bug?
The same prompt works normally with GPT-5.6 Sol.
Session ID: 01a0d259-2dda-7ae2-aa88-24319f53dd8d
Occurred at: 2026-09-24 15:39 GMT+8
Reproduction rate: 100%
What is the expected behavior?
GPT-6 Astra should process harmless prompts such as “输出 1-10” normally. The prompt should not be classified as a potential usage-policy violation, and the user should not be forced to switch models.
Additional information
Account: Personal ChatGPT Pro 5x
Selected model: GPT-6 Astra
Codex Desktop version: 26.917.71314
Platform: macOS
Affected surfaces: Codex Desktop and ChatGPT Work on the web
Minimal reproducer: 输出 1-10
Control test: The same prompt works with GPT-5.6 Sol.
Network configurations tested:
The issue reproduces on both network configurations. GPT-6 Astra worked during the first few weeks after launch and stopped working approximately one week ago.
Possibly related: #46753 and #45339.