OpenAI's new cyber model found 81 Firefox flaws

OpenAI's GPT-6 Cyber reportedly found 81 new high-severity Firefox flaws. Access is restricted.

According to BleepingComputer, reporting on OpenAI's announcement Thursday, GPT-6 Cyber is a model OpenAI says improves on earlier ones at finding vulnerabilities, validating their impact and helping developers create fixes. In testing, it found 81 previously unknown high-severity Firefox flaws and helped develop patches, many already released by Mozilla.

In separate Linux kernel research, the model identified 2,308 potential flaws, reproduced 1,724 vulnerabilities, and developed local privilege escalation exploits for 948 kernel vulnerabilities. On the ExploitGym benchmark it exploited 48.5% of vulnerabilities, versus 32.1% for GPT-5.6 Cyber. OpenAI classifies it as 'Critical' under its Preparedness Framework, up from 'High'.

The model is not public. Access is limited to approved customers in Daybreak Red, with wider Daybreak Red access planned soon. OpenAI warns defenders have a limited window to use such models to patch flaws before similar capabilities reach attackers. Figures are as reported and not independently verified.

“It is clear that we have a limited window to use frontier models to secure software,”OpenAI, statement via BleepingComputer

The facts we checked

According to BleepingComputer, reporting on OpenAI's announcement. Figures are as reported and not independently verified.

1. OpenAI revealed GPT-6 Cyber, an AI model it says significantly improves on earlier models at finding software vulnerabilities, validating their impact and helping developers create fixes. The company announced it on Thursday.Reported · BleepingComputer
2. In security testing, GPT-6 Cyber discovered 81 previously unknown high-severity vulnerabilities in Firefox and helped develop patches, many of which Mozilla has already released.Reported · BleepingComputer
3. In separate Linux kernel research, the model identified 2,308 potential security flaws and successfully reproduced 1,724 vulnerabilities.Figure · BleepingComputer
4. The model developed local privilege escalation exploits for 948 distinct Linux kernel vulnerabilities, which could allow an unprivileged user to gain root access.Figure · BleepingComputer
5. In benchmark testing, GPT-6 Cyber exploited 48.5% of vulnerabilities in ExploitGym, compared with 32.1% for GPT-5.6 Cyber.Figure · BleepingComputer
6. OpenAI classified GPT-6 Cyber as having 'Critical' cybersecurity capabilities under its Preparedness Framework, up from the 'High' classification given to its previous model.Reported · BleepingComputer
7. The model is not publicly available. Access is restricted to a select group of approved customers in OpenAI's Daybreak Red, and OpenAI plans to expand access to all Daybreak Red customers soon.Reported · BleepingComputer
8. Daybreak Red supports advanced exploit development and authorized security testing. Daybreak Blue lets verified defenders investigate vulnerabilities and validate fixes with fewer restrictions.Reported · BleepingComputer
9. OpenAI warned that defenders have a limited window to use increasingly powerful AI models to find and patch vulnerabilities before similar capabilities become more widely available to attackers.Reported · BleepingComputer
10. "It is clear that we have a limited window to use frontier models to secure software," OpenAI said.Quote · BleepingComputer
11. OpenAI said it plans to expand its Daybreak programs to organizations of all sizes, independent researchers and open-source maintainers.Reported · BleepingComputer
12. BleepingComputer reports that Chinese hackers used an AI-powered penetration testing framework called ARTEX to automate attacks against more than 200 organizations worldwide, compromising at least nine.Reported · BleepingComputer

Sources

• BleepingComputerOpenAI's new GPT-6 Cyber AI finds 81 Firefox flaws, 948 Linux bugs