OpenAI posted a PR scans demo for pull request workflows.
Before that, Codex Security centered on the openai/codex-security CLI and TypeScript SDK. Those tools find, investigate and fix code vulnerabilities. The demo moves that check into the pull request view with a presenter reviewing results.
The shift is small and practical.
In GitHub, Codex flags only P0 and P1 issues, per ChatGPT Learn docs. That filter keeps review comments on high priority risks and hides lower severity notes.
The demo page lists no pricing, limits or rollout date.





