OpenAI has reportedly said that its security-focused model, GPT-6 Cyber, identified 81 high-severity bugs in Mozilla Firefox and produced working exploits for 948 flaws in the Linux kernel, with access limited to vetted red teams.
OpenAI · AI security research · Reported claims
GPT-6 Cyber: reported bug and exploit counts from a restricted offensive-security model
OpenAI has reportedly said its security model found 81 high-severity Firefox bugs and produced working exploits for 948 Linux kernel flaws. Access is limited to vetted red teams. The figures are company claims that have not yet been independently verified.
81
High-severity bugs found in Mozilla Firefox
948
Linux kernel flaws with reported working exploits
0.07%
Reported out-of-scope actions in internal tests (test count not stated)
Scale comparison: findings per target
Each block = 10 flaws. Column height is proportional to the count.
81
Firefox
948
Linux kernel flaws
The Linux figure is roughly 11.7 times the Firefox figure. The two counts measure different things and cover different codebases, so the comparison shows scale, not relative difficulty.
Why it could help defenders
Could shorten the time from discovery to patch.
Findings reach defenders before attackers can use them.
Adds a concrete data point to the debate on automated exploit generation.
Why it draws caution
The same capability could be misused if it reached attackers.
Restricted access is the main safeguard described.
Security researchers still debate whether it helps defenders more than attackers.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...