OpenAI Rolls Out GPT-5.5-Cyber to 'Critical Cyber Defenders' for Cybersecurity Support

OpenAI has prepared a new frontier cybersecurity model, GPT-5.5-Cyber. The model, designed for 'critical cyber defenders', will be available only to a select group of vetted professionals and institutions for improving cyber defenses. OpenAI will collaborate with governments and the wider ecosystem to grant access.

OpenAI announced the rollout of GPT-5.5-Cyber, a specialized frontier AI model designed to support defensive cybersecurity operations . Launched in late April 2026, the model is restricted to vetted professionals and institutions through OpenAI's Trusted Access Program — a framework that screens participants for legitimate defensive security use cases in critical infrastructure and sensitive environments.

GPT-5.5-Cyber is classified at the 'High capability' tier in the cybersecurity domain, capable of identifying vulnerabilities, analyzing threat patterns, and supporting incident response. OpenAI has implemented tighter safety controls against higher-risk activities, including protections against repeated misuse, and the model does not reach the 'Critical' threshold that would enable autonomous zero-day exploit development against hardened systems. The company plans to collaborate with governments and the wider ecosystem to refine deployment protocols.

The rollout reflects a broader industry dynamic: frontier AI models are increasingly being applied to specialized, high-stakes domains where human capacity constraints are most acute. Cybersecurity, where the defender-to-attacker imbalance is chronic, is a natural fit. By maintaining restricted access through a vetted program, OpenAI is attempting to capture the defensive benefits of the model while limiting the risk of capability transfer to malicious actors — an approach that will be closely watched as a template for responsible AI deployment in sensitive domains.