OpenAI has put Codex Security into research preview, framing it as an app security agent that finds, validates, and patches vulnerabilities with deeper system context. The pitch: fewer false positives, higher-confidence fixes, and a tighter path from review to remediation.