GPT-6 Cyber: Cybersecurity model for vulnerability investigation and fix verification
In brief
OpenAI's GPT-6 Cyber model investigates vulnerabilities, reproduces them in authorized test environments, and verifies fixes. It is trained to stay within authorized scope.
How it works
You access GPT-6 Cyber via the OpenAI API after completing Daybreak setup. The model scans codebases, identifies vulnerabilities, reproduces them in a controlled environment, and verifies patches. It operates only within authorized boundaries.
Security teams and developers building automated vulnerability detection and remediation workflows.
Availability
Sources
5 merged into this card- 1GPT‑6 Cyber: Advancing the frontier of aligned cybersecurity agentsAnnouncement · openai.com · 05.11
- 2Use Daybreak in the Responses APIHow-to guide · developers.openai.com
- 3GPT-6 Cyber ModelHow-to guide · developers.openai.com
- 4OpenAI API changelog — November 5, 2026Release notes · platform.openai.com · 05.11
- 5Codex Security: From Vulnerability to Verified Fix | DevDay 2026Video · youtube.com · 11.11
More from OpenAI
Codex TUI WebRTC Realtime Voice Calls
Codex v0.170.0 adds WebRTC negotiation, allowing users to conduct realtime voice calls directly in the terminal user interface.
Codex: Refreshable OpenTelemetry HTTP Headers
Codex now automatically refreshes authentication headers for OpenTelemetry HTTP exporters, ensuring continuous telemetry without token expiration issues.