Details
- OpenAI announces that GPT-5.6 Sol has set a new state of the art in cybersecurity on The Last Ones cyber range, a benchmark environment for testing real-world security performance.
- The company reports that GPT-5.6 Sol’s capabilities are already translating into defensive outcomes, helping teams find, validate, and fix vulnerabilities in production code rather than only synthetic test cases.
- This capability is being productized through Codex Security, a plugin in the Codex developer platform that lets users run automated security scans on their codebases using GPT-5.6 Sol.
- To get started, users add the Codex Security plugin inside Codex; once installation completes, the interface exposes a Try in chat button to initiate a new Codex chat session.
- Clicking Try in chat launches a pre-configured Codex Security scan prompt, after which users choose a specific folder or repository for analysis, allowing the model to search for vulnerabilities and suggest fixes.
- GPT-5.6 Sol is OpenAI’s highest-rated cybersecurity model, achieving state-of-the-art performance on exploit benchmarks while being tuned to favor defensive tasks such as vulnerability identification and patch development over end-to-end offensive attacks.
- Codex Security operationalizes these capabilities for software teams, turning frontier-model security reasoning into a repeatable workflow that can be integrated into continuous integration and code review pipelines.
- The launch builds on OpenAI’s broader GPT-5.6 family, which emphasizes stronger cyber safety layers and controlled access, and extends those advances directly into developer tooling via Codex.
Impact
By tying GPT-5.6 Sol’s state-of-the-art cybersecurity performance directly into the Codex Security plugin, OpenAI is moving frontier security capabilities closer to everyday developer workflows. This pressures rival AI platforms to offer similarly integrated defensive tooling and may accelerate adoption of AI-assisted secure coding practices, particularly among teams that lack dedicated security engineers.
