OpenAI announced on Tuesday the launch of GPT-5.6 Cyber, an artificial intelligence (AI) model specialized in cybersecurity. This model is tailored for advanced vulnerability research, exploit validation, and security testing.
In ExploitGym, which evaluates the ability to convert known vulnerabilities into actionable attack code, GPT-5.6 Cyber achieved the highest success rate among all models OpenAI has tested to date. It also completed 95% of requests in advanced cybersecurity task assessments.
Beyond research benchmarks, GPT-5.6 Cyber has been applied to study real-world software vulnerabilities.
OpenAI revealed that while examining V8, Chrome’s JavaScript engine, the model uncovered two previously unknown vulnerabilities that could be exploited in tandem. After verification, the OpenAI research team reported these findings to Google following established vulnerability disclosure protocols, and Google promptly patched the issues.
Alongside GPT-5.6 Cyber’s release, OpenAI is introducing the ‘Daybreak’ program, offering cybersecurity models to individuals and organizations for approved defensive operations. The program is split into Daybreak Blue and Daybreak Red.
Daybreak Blue, based on GPT-5.6 Sol, is an entry-level tier with safety measures tailored for approved defensive tasks. OpenAI recommends most security teams start with Daybreak Blue.
Daybreak Red, built on GPT-5.6 Cyber, caters to organizations engaged in more specialized security work, including advanced vulnerability research, exploit development, and red team activities.
Both tiers are exclusively available to pre-approved individuals and organizations conducting authorized security operations. Strict controls are in place, including identity verification, account security, monitoring, usage restrictions, and legal compliance checks.
Access to Daybreak Blue and Red is granted to those who have completed Daybreak registration and are qualified Amazon Bedrock users. They can utilize these models within their existing AWS security and governance frameworks.
Furthermore, OpenAI is expanding its Daybreak Cyber Partner Program, collaborating with major security and service providers such as Accenture, Cisco, Cloudflare, CrowdStrike,
IBM, and Palo Alto Networks’ Unit 42.
These partners can integrate OpenAI’s cutting-edge cybersecurity model into their products, managed security services, and client projects. This enables more organizations to leverage advanced cybersecurity capabilities without the need to develop and manage complex cybersecurity AI programs internally.