GPT-6 Cyber has been created thinking about cybersecurity and has been used not only to identify vulnerabilities in other companies' systems, but also to strengthen OpenAI's own infrastructure.
© Getty Images
07/11/2026
by
A OpenAI presented a new model of Artificial Intelligence created thinking about cybersecurity. It is called GPT-6 Cyber and, according to OpenAI, it is better to find and develop corrections for security vulnerabilities than the previous model-
The ChatGPT company also explains that during the development of GPT-6 Cyber, the Artificial Intelligence model managed to discover 81 new security failures in the Firefox browser, which were categorized by Mozilla as being very serious.
Besides having tested GPT-6 Cyber in other environments, OpenAI also trained the Artificial Intelligence model on its own servers, taking advantage of situations where agents managed to escape from controlled testing environments to reinforce security protections and prevent this from happening again in the future.
“ We trained GPT-6. Cyber for advanced cybersecurity work with improved alignment”, can be read in the publication shared by one of the elements of OpenAI, Xiangyu Qi, on their social network page X. “ It has been fueling our security work for some time and has been behind many of our recent security highlights: finding and helping to correct vulnerabilities with high gravity in Firefox, developing ‘exploits’, presenting corrections for vulnerabilities in the Linux kernel and strengthening OpenAI’s own infrastructure.”.
Despite the security guarantees shared by OpenAI, GPT-6 Cyber will remain unavailable to the general public. OpenAI has classified this model as “critical” with regard to cybersecurity and, for the time being, will allow only a small group of cybersecurity teams to have access to the new model of Artificial Intelligence.
Introducing GPT-6 Cyber! We trained it for advanced cybersecurity work with improved alignment. It has been powering our security work for settle a while and has been behind many of our recent security highlights: finding and helping patch high-severity vulnerability in Firefox, developing exploits and proposing patches for Linux kernel vulnerability, and strengthening OpenAI’s own infrastructure through red-teaming. Today, we’re rolling it out to Daybreak Red to keep the cyber frontier in defenses’ hands. https://t.co/BHDIt8O6Ck
— Xiangyu Qi (@xiangyuqi pton) November 5, 2026