OpenAI launches GPT-6 Cyber security model

OpenAI has introduced GPT-6 Cyber, a model focused on cybersecurity work. According to the company, the system is designed to help find exploitable vulnerabilities and develop fixes, with an emphasis on staying within an authorized scope. OpenAI says the model has been trained for authorized security work, aiming to help organizations identify weaknesses in their systems and remediate them. The launch comes as AI laboratories face increasing scrutiny over their role in protecting digital infrastructure and preventing misuse. Security teams have long used automation to scan code and networks, but generative AI models could change how vulnerabilities are discovered, prioritized and fixed. A model built specifically for this area could help defenders move faster, provided it is used responsibly and with clear permission boundaries. OpenAI’s framing is notable. The company is positioning GPT-6 Cyber around authorized use rather than unrestricted offensive testing. That distinction matters because vulnerability research can be dual-use: the same techniques that help defenders find flaws could be misused if applied without consent. By emphasizing authorized scope, OpenAI is trying to draw a line between legitimate security assessments and improper activity. The announcement does not provide a full technical breakdown in the summary, so questions remain about how the model performs on real-world codebases, how it handles sensitive systems and what safeguards accompany access. Those details will be important for security leaders evaluating whether to adopt it. Still, the release signals a growing push to tailor advanced AI models for specialized professional domains, with cybersecurity as one of the most consequential.