OpenAI GPT-5.5-Cyber: New AI for Cybersecurity Defenders

OpenAI has just launched GPT-5.5-Cyber, a new AI model specifically for cybersecurity defenders. This specialized AI scored an impressive 81.9% on the CyberGym benchmark. It even successfully completed simulated corporate cyberattacks. What's interesting is that the UK AISI found it nearly as capable as Anthropic's Claude Mythos, succeeding in 20% of a 32-step attack simulation, compared to Mythos's 30%. Here's the thing: while Mythos is locked to about 40 organizations, OpenAI is offering tiered access through its Trusted Access for Cyber program. This model is designed to be more permissive for security tasks, helping legitimate defensive work without triggering safety guardrails. It can even generate vulnerability exploitation plans and then validate them with simulated attacks. This means cybersecurity professionals now have another powerful AI tool to help protect against digital threats.