OpenAI Launches GPT-5.5 and GPT-5.5-Cyber for Cybersecurity Defenders

OpenAI has released GPT-5.5, its most advanced model, along with a specialized variant GPT-5.5-Cyber, aimed at empowering cybersecurity defenders. The models are part of a broader initiative to democratize AI-powered defense, as outlined in the action plan 'Cybersecurity in the Intelligence Age'. GPT-5.5 with Trusted Access for Cyber (TAC) provides verified defenders with reduced refusals for defensive tasks like vulnerability triage, malware analysis, and patch validation, while blocking malicious activities. GPT-5.5-Cyber, available in limited preview, offers more permissive behavior for specialized workflows such as red teaming and penetration testing, under stronger verification and monitoring. OpenAI is partnering with security vendors including Cisco, Intel, SentinelOne, Snyk, Gen Digital, Semgrep, and Socket to integrate these capabilities into real-world workflows. The models aim to accelerate the security flywheel by improving vulnerability research, detection, software supply chain security, and network enforcement. Access to TAC requires phishing-resistant authentication, with Advanced Account Security mandatory from June 1, 2026. Codex Security is also being offered to open-source maintainers for vulnerability identification and patching. During alpha testing, GPT-5.5-Cyber has been used for automated red-teaming of critical systems.