GPT-5.5 -Cyber is out in «limited preview,» available to vetted defenders


Two models models are launching today; GPT-5.5 with «Trusted Access for Cyber» that requires some vetting to get into. It can handle defensive security, do code review, malware analysis and patch validation.

GPT-.5.5-Cyber requires stronger verification and does specialized workflows, red teaming, penetration testing and controlled validation.

GPT-5.5 Cyber was earlier found on par with Anthropic’s Mythos model, that has been spooking the establishment lately.

The vetting approach for the Cyber model has been «informed by conversations with cybersecurity and national security leaders across federal and state government and major commercial entities,» OpenAI says.

Read more: OpenAI’s blog, CNBC, Axios.