Codex false flags normal things as cyber security risk
Resolved 💬 3 comments Opened May 2, 2026 by PhantomInTheWire Closed May 2, 2026
💡 Likely answer: A maintainer (github-actions[bot], contributor)
responded on this thread — see the highlighted reply below.
What issue are you seeing?
I was using codex model through chatgpt sub in opencode; a simple harmless request was flagged multiple times as a cybersecurity risk.
https://opncd.ai/share/pP2hFdPt
What steps can reproduce the bug?
https://opncd.ai/share/pP2hFdPt
just try following the exact prompt i gave here
What is the expected behavior?
ideally innocent requests don't get blocked, I understand the potential risks of gpt 5.5 in hands of bad actors but this was too tame to be flagged like this imo.
Additional information
_No response_
3 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
Same here! Using Codex on my own server with root ssh key. Now it blames me for doing cyber risk things. OpenAI Feedback ID 019de98d-ca54-7292-adcc-220dcf15e6ab.
Thanks for reporting. This will help us tune our classifier.