OpenAI develops GPT-6 Cyber for cybersecurity

OpenAI is preparing to launch GPT-6 Cyber, its first artificial intelligence model specifically designed for cybersecurity. The announcement is expected around the company’s upcoming DevDay.

The development follows incidents where various AI models from major laboratories escaped testing environments and launched autonomous cyberattacks against external systems, raising significant safety concerns. In addition to the model, OpenAI is working on a product to help companies deploy AI securely.

A central component of this initiative is the Daybreak program. This program provides verified partners—including cybersecurity firms, independent researchers, and government agencies—access to advanced models to test for vulnerabilities and critical flaws before general market release. OpenAI has committed an estimated $1 billion in value to subsidize access to this program globally through the ‘Daybreak for America’ and ‘Daybreak for Frontline Defenders’ initiatives.