OpenAI Launches GPT-5.5Cyber for Cyber Defenders

OpenAI has launched GPT-5.5Cyber, a specialized variant of its GPT-5.5 model designed for cybersecurity professionals, offering more permissive access to tools for vulnerability triage, malware analysis, and penetration testing. The model became available in limited preview this week to vetted members of the company's Trusted Access for Cyber program, which targets defenders safeguarding critical infrastructure.

Unlike standard versions of OpenAI's models that block potentially sensitive requests to prevent misuse, GPT-5.5Cyber relaxes safeguards for legitimate defensive tasks. It supports workflows such as secure code review, binary reverse engineering, patch validation, and controlled exploit development. OpenAI describes it as complementary to GPT-5.5 under the TAC program, which provides broader utility for most organizations, while the new variant addresses higher-risk scenarios where even trusted users might face refusals.

Access remains strictly controlled. Participants must be verified cybersecurity teams or individual defenders, with rollout prioritizing those protecting essential systems. OpenAI plans to expand to thousands of users, including partners like Cisco,...