OpenAI Launches GPT-5.5-Cyber for Security Teams

OpenAI announced the limited preview of GPT-5.5-Cyber, a specialized AI model tailored for cybersecurity teams. This version is designed to enhance workflows such as vulnerability identification and malware analysis by operating with fewer restrictions than the general GPT-5.5 model. This rollout occurs one month after rival Anthropic introduced its Claude Mythos Preview. The announcement highlights the intensifying competition between major AI firms in providing advanced tools for the cybersecurity sector.

OpenAI has rolled out GPT-5.5-Cyber, a specialized AI model variant, to vetted cybersecurity teams, marking a competitive move one month after Anthropic debuted its Claude Mythos Preview.

GPT-5.5-Cyber Capabilities

According to a blog post, the GPT-5.5-Cyber preview is designed not necessarily as a massive leap in raw cyber capability, but rather to operate with greater permissiveness for security-related tasks. This specialized version allows vetted teams to perform complex workflows that were previously restricted by the safeguards built into the generally available GPT-5.5 model.

Key functions available to participating teams include:

Vulnerability identification and triage.
Patch validation.
Malware analysis.

OpenAI stated that this specialized access allows a smaller group of partners to study advanced workflows where specific access behaviors are critical.