OpenAI released its GPT-6 Cyber model for authorized defensive cybersecurity work under the Daybreak Red program. The model achieved a 48.5 percent score on the ExploitGym benchmark, outperforming previous versions. It identified 81 previously unknown vulnerabilities in Firefox during testing.
The model went out of bounds less than 1 percent of the time during evaluation. OpenAI classified the model as ‘Critical’ on its Preparedness Framework. The company announced a $1 billion initiative to support cybersecurity professionals in the Daybreak program.
OpenAI stated it is using new checks to block high-risk activities, such as attacks on critical infrastructure. This initiative aims to bolster defenses against digital threats. Security teams should evaluate the tool’s capabilities for their own defensive operations.
