OpenAI says an unreleased cyber-focused model identified, validated, and patched 948 Linux kernel privilege escalation flaws in a single autonomous run, at a total API cost of $1.2 million.
OpenAI · Autonomous cyber model · Linux kernel
One autonomous run, 948 kernel privilege escalation flaws addressed
OpenAI says an unreleased cyber-focused model found, validated, and patched 948 Linux kernel privilege escalation flaws in a single continuous session, at a reported API cost of $1.2 million. The figures are the company's own and have not yet been independently verified.
948
Kernel privilege escalation flaws reported as addressed
$1.2M
Total reported API cost of the run
~$1,270
Implied API cost per flaw ($1.2M ÷ 948)
Scale of the claim, drawn to size
Each block represents roughly 50 flaws. The full column equals the 948 reported in the single run.
948
Flaws in one autonomous session, covering the full cycle of finding, confirming, and patching.
Severity of each flaw and whether patches were accepted upstream are not yet independently confirmed.
How the single run is described to work
STEP 1
Find
Identify candidate flaws in kernel code
→
STEP 2
Validate
Confirm each candidate is a real flaw
→
STEP 3
Patch
Produce fixes for the confirmed flaws
Supporters see
A faster way to shrink the backlog of latent kernel bugs
AI applied to defensive vulnerability research at scale
Continuous auditing that may cost less than manual review
Critics and observers note
Figures come from OpenAI itself and are not independently verified
The same capability could locate exploitable flaws offensively
Severity and upstream patch acceptance remain unconfirmed
What decides the claim
The headline holds only if the patches survive outside review.
Merged upstream
How many patches reach mainline kernel releases
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...