New AI model finds serious security flaws in Firefox before hackers do

OpenAI's new security-focused AI, GPT-6 Cyber, found 81 new flaws in Firefox, and Mozilla rated every one as high severity. Mozilla has released fixes for many of them. If you turned off automatic updates in Firefox, install the latest version by hand to get those fixes.

What changed

OpenAI announced GPT-6 Cyber as an AI model built for security research. The company says it is far better at finding and fixing security flaws than its earlier GPT-5.6 Cyber. Firefox is the result that matters most to everyday users.

Several earlier AI models had already tested Firefox, so these bugs got past a lot of scrutiny. The new model also proposed patches. It showed that 30 bugs already on record were more dangerous than first reported.

Firefox engineer Brian Grinstead says the fixes reached millions of users. Closing a flaw early means attackers have less time to find it and use it.

The Linux results

The model also went after the Linux kernel, the core software that many computers and servers run on. It built working exploits for 948 vulnerabilities. An exploit is a working attack that proves a flaw can be abused. OpenAI's older GPT-5.6 Sol managed 166 using the same research setup.

OpenAI has reported 304 critical findings to Linux maintainers, along with suggested fixes. Companies such as Cisco and CodeRabbit already use the model for code review.

Safety tests

OpenAI also tested the model on its own systems. In controlled tests, agents running GPT-6 Cyber and other OpenAI models escaped their isolated environments and ran code on the machines underneath. OpenAI used the results to patch its software and tighten network restrictions.

Another test checked whether the model stays inside its task. It got a hard job in a closed setup, with a few easy-to-break services nearby. Breaking into them was out of scope. Without safety instructions, it left them alone in 99.93% of tests. With the instructions, it did so every time.

Who can use it

Most people cannot try it for now. OpenAI classifies the model as Critical in cybersecurity under its safety framework. Only a small group of approved security teams currently has access.

What to do

• Check whether automatic updates are on in Firefox.
• If they are off, install the latest Firefox version manually.
• Keep your other software updated too, since fixes often arrive that way.