OpenAI Announces GPT-6 Cyber Model for Vulnerability Detection and Fixes

What Happened

OpenAI has announced GPT-6 Cyber, a new cybersecurity AI model designed to identify exploitable vulnerabilities and develop automated fixes [1, 2]. The model is trained to operate within an authorized scope [1]. According to OpenAI, the model is classified as “Critical” under its Preparedness Framework, meaning it can identify and develop functional zero-day exploits of all severity levels in many hardened real-world critical systems without human intervention [2]. This is an upgrade from GPT-5.6 Cyber, which was classified as “High” [2].

In testing using the ExploitGym HoneyPot evaluation, OpenAI reported that GPT-6 Cyber went out of bounds less than 1 percent of the time, compared to approximately 5 percent with GPT-5.6 Cyber [2]. Fouad Matin, head of OpenAI’s cyber preparedness and safety team, stated during a press briefing that the model performs its requested tasks over 99 percent of the time [2]. On the ExploitGym benchmark, GPT-6 Cyber scored 48.5 percent, which is 16 points higher than GPT-5.6 Cyber’s 32.1 percent and a few points higher than GPT-6 Astra’s 42.4 percent [2].

OpenAI reported several specific uses for the model:

• It helped find 81 previously unknown vulnerabilities in the Firefox web browser [2, 3].
• It was used to scan and prepare patches for the Linux kernel, identifying 948 potential issues [3].
• It has been used to help harden OpenAI’s own training infrastructure [2].

The model is currently rolling out to select members of the Daybreak Red program, which is for companies and organizations performing authorized defensive cybersecurity work [2]. OpenAI plans to expand access to all Daybreak Red customers in the coming weeks [2]. Additionally, OpenAI announced a $1 billion initiative to support cybersecurity professionals in the program who work on critical infrastructure, such as water systems or the electrical grid [3].

Who Is Affected

• Daybreak Red members: These are companies and organizations using models for authorized defensive cybersecurity work; they will receive expanded access in the coming weeks [2].
• Cybersecurity professionals: OpenAI is launching a $1 billion initiative to support those working on critical infrastructure [3].
• Open-source maintainers: OpenAI stated its goal is to bring advanced security research to independent researchers and open-source maintainers [3].