Key Takeaways
- OpenAI's 948 Linux kernel vulnerabilities: OpenAI said on Nov. 4 that its upcoming cyber model built 948 working Linux kernel root exploits in one run.
- Cost: The run cost $1,213,430 in API credits, or $1,279 per exploit, about 59% less per exploit than a July run that found 166.
- Patching: OpenAI generated candidate patches and is reporting the flaws to Linux maintainers with its Patch the Planet partners.
- Defenders: Cloudflare the same day urged developers to audit their own code with its open-source security audit skill for coding agents.
OpenAI said on Nov. 4, 2026, that a pipeline running its upcoming cyber model validated and generated patches for 948 local privilege escalation vulnerabilities in the Linux kernel, at an API cost of $1.2 million. Each flaw could let an unprivileged local user gain root access, the highest level of control over a Linux system.
How OpenAI's Pipeline Found 948 Linux Kernel Vulnerabilities
In the mid-October run, the pipeline identified 2,308 candidate findings, dynamically reproduced 1,724 vulnerabilities and produced 948 working exploits, along with proofs of concept for 291 information leak flaws, according to OpenAI's write-up. The test machines ran a stock desktop operating system, not a deliberately weakened target.
The pipeline covered vulnerability discovery, dynamic validation, exploitability assessment and patch generation, and human researchers reviewed the results before reporting them. Its scout agents selected 5,034 audit targets covering about 2.8 million lines, roughly 7% of a kernel with more than 40 million lines of code, Unite.AI reported. OpenAI said that, to the best of its knowledge, the agents discovered and used at least 10 novel Linux exploitation techniques. It did not name the model, which SecurityWeek describes as unreleased.
The Cost per Linux Root Exploit Fell About 59% Since July
The October run cost $1,213,430, or about $1,279 per exploit, SecurityWeek reported. An earlier scan in late July, powered by GPT-5.6 Sol, yielded 166 unique exploits at a cost of $513,690. That works out to about $3,095 per exploit, so the cost per exploit fell about 59% in roughly three months while the number found rose 5.7 times, despite three months of kernel hardening between the scans.
By OpenAI's math, at the $20,000 and $70,000 that the Pwn2Own and TyphoonPWN contests have historically paid for Linux kernel privilege escalation exploits, the 948 bugs would be worth $18.9 million to $66.3 million, 16 to 55 times the scan's cost. OpenAI presents its costs as upper bounds, because the research focused on method rather than cost.
What 948 Linux Kernel Flaws Mean for Patching and Defenders
Maintainers were already moving. "Within 10 days while the scan was running, we observed over 100 Linux LPEs had been patched or had staging patches from other contributors, with ~300 more being publicly discussed," OpenAI wrote. OpenAI has sent hundreds of reports and patches for critical vulnerabilities and plans to send the rest gradually through Patch the Planet, an initiative it introduced with Trail of Bits on June 22, 2026.
OpenAI also warned administrators. "For the near and medium term, having access to a low-privileged account on an operating system, that isn’t up-to-date on its latest available patches should be considered functionally equivalent to root access for a well-resourced threat actor."
The same day, Cloudflare's developer account pointed defenders to a free tool. "With the recent wave of vulnerability disclosures, it’s a good time to take a closer look at your own code," Cloudflare Developers wrote in a post on X. The post links the open-source security audit skill that seeded Cloudflare's own vulnerability discovery harness. It runs six phases: map the codebase, hunt for vulnerabilities, challenge each candidate, record the findings, verify them and report the results.
OpenAI wrote that prioritization and validation, not discovery, remain the operational bottlenecks. Cloudflare's June blog post makes a similar case: better context cut its harness's initial validation rejection rate from 40% to 11%, while the share of high-integrity findings climbed from 35% to 58%, representing about 12,057 lifetime findings.
People Also Ask
How many Linux kernel vulnerabilities did OpenAI's AI find?
OpenAI's pipeline reproduced 1,724 Linux kernel vulnerabilities and built 948 working local privilege escalation exploits in a mid-October 2026 run, OpenAI said on Nov. 4, 2026. The run started from 2,308 candidate findings.
How much did OpenAI's Linux kernel scan cost?
OpenAI's Linux kernel run in mid-October 2026 cost $1,213,430 in API credits, or about $1,279 per working exploit. OpenAI describes the figures as upper bounds, because the research measured method and model capability rather than cost.
What is a local privilege escalation vulnerability?
A local privilege escalation vulnerability lets someone with a low-privileged account on a machine gain higher rights. Each of the 948 Linux kernel flaws in OpenAI's run could let an unprivileged local user gain root access.
What is Cloudflare's security audit skill?
Cloudflare's security audit skill is an open-source tool on GitHub that turns a coding agent into a security auditor in six phases. Cloudflare says the skill seeded its own vulnerability discovery harness.
Sources: OpenAI, OpenAI on X, SecurityWeek, Unite.AI, Cloudflare Developers on X, Cloudflare Blog, GitHub. Related: OpenAI profile.