Finding a vulnerability is only one part of security work: teams must also validate its impact and test a fix. OpenAI’s November 5 announcement introduces GPT-6 Cyber, a model designed for authorized security research and remediation.
What is it used for?
OpenAI says the model helps investigate vulnerabilities, validate impact and propose fixes. In its controlled ExploitGym evaluation, GPT-6 Cyber scored 48.5%, compared with 32.1% for GPT-5.6 Cyber. These are results under the company’s test conditions, not a guarantee for every real-world environment.
OpenAI says the model produced evidence and reproducible tests for 81 new Firefox vulnerabilities classified by Mozilla as high severity, and that it disclosed the findings to Mozilla. In Linux research it reproduced 1,724 vulnerabilities and developed working exploits for 948.
Access and boundaries
GPT-6 Cyber is initially available through the API and Codex to a select group of approved Daybreak Red customers. OpenAI plans to expand access to more Daybreak Red customers. It reports dedicated evaluations and monitoring for staying within authorized scope, and recommends sandboxing and human oversight for high-risk workflows.
