OpenAI has released GPT-6 Cyber, an advanced AI model rated 'Critical' for cybersecurity, aimed at helping security experts identify and patch system vulnerabilities before malicious actors can exploit them.
- GPT-6 Cyber rated 'Critical' for cybersecurity under OpenAI's safety framework.
- Used to discover hundreds of previously unknown vulnerabilities in Firefox and Linux.
- Access limited to vetted customers through OpenAI's Daybreak Red program.
What happened
OpenAI has launched GPT-6 Cyber, an AI model engineered to assist cybersecurity professionals in detecting and exploiting software vulnerabilities. The model was released to a small, vetted group of users under OpenAI's Daybreak Red program, which provides advanced access for authorized security research such as penetration testing and exploit creation.
This latest iteration of OpenAI’s cybersecurity-focused AI carries a 'Critical' rating under the company’s own safety framework, an upgrade from the 'High' rating assigned to the previous GPT-5.6 Cyber model. OpenAI's internal research demonstrated the model's capabilities by identifying 81 new high-severity Firefox vulnerabilities and developing exploits for nearly 1,000 Linux kernel vulnerabilities, many previously unknown.
Why it matters
The release of GPT-6 Cyber marks a significant development in AI-assisted cybersecurity by providing researchers with a powerful tool to proactively identify and mitigate risks before they can be leveraged by malicious actors. Its critical rating underscores both the model’s potency and the necessity of stringent controls on who can access such advanced technology.
With cyber threats becoming more sophisticated and the misuse of AI growing, GPT-6 Cyber provides a dual benefit: accelerating vulnerability research while enabling timely patching of critical security flaws. This is particularly important for software widely used in enterprise environments, such as Mozilla’s Firefox browser and the Linux operating system kernel.
What to watch next
OpenAI plans to gradually expand GPT-6 Cyber access to all approved customers within its Daybreak Red tier over the coming weeks, potentially increasing its impact on the cybersecurity ecosystem. Meanwhile, OpenAI has also made the prior GPT-5.6 Cyber model available to all Daybreak Red users, broadening the available tools for security research.
Observers should monitor how this technology influences vulnerability discovery rates and patch development timelines across major open-source projects. Additionally, the broader implications of releasing such powerful AI tools, including potential misuse and regulatory responses, will be key areas to watch in the evolving landscape of AI and cybersecurity.