Model
113d agoOpenAI Releases GPT-5.5-Cyber For Security Experts
GPT-5.5-Cyber executes active exploits against test servers and bypasses standard security refusals. OpenAI limits access to verified infrastructure defenders, including CrowdStrike and Cisco. This model directly challenges Anthropic's Mythos Preview.
The Signal
Security practitioners now have a tool specifically tuned for offensive testing and vulnerability discovery.