How much did OpenAI's Linux kernel scan cost?

OpenAI reports the mid-October 2026 Linux kernel campaign cost $1,213,430 in API credits end to end. The largest stage was exploitation at $775,929; candidate discovery cost $316,549 and validation $73,330. A successful exploit averaged $359. OpenAI calls these figures an upper bound, since better harness engineering could cut the cost of each stage.

Have the Linux kernel vulnerabilities OpenAI found been fixed?

Partly. When OpenAI's scan and manual analysis finished after 10 days, 61 of the local privilege escalation bugs were already patched upstream, 43 had staging patches and 299 were being reported and discussed publicly. Following the Linux kernel community's advice, OpenAI started with critical vulnerabilities and is sending the rest gradually with Patch the Planet partners.

How does the new OpenAI cyber model compare with GPT-5.6 Sol on the kernel?

In a late-July 2026 scan with the same research harness, GPT-5.6 Sol found and exploited 166 unique Linux kernel privilege escalation bugs for $513,690 in API credits. The mid-October run with OpenAI's newer, unreleased cyber model found and exploited 948 for $1,213,430. OpenAI's takeaway is to rescan audited code whenever model capability improves.

Did OpenAI's pipeline work against Google's hardened kernelCTF target?

Yes. Run under Google's kernelCTF environment, a hardened kernel with a reduced attack surface, OpenAI's pipeline produced 25 information-leak proofs of concept and 21 local privilege escalation exploits over 112 hours for about $32,936 in API credits. OpenAI estimates those 21 would be worth about $1.5M in bounties if accepted, roughly 45x the cost.

OpenAI · 2026-11-04 · major

A thousand fewer paths to root — OpenAI agents find 948 Linux root exploits

In short

OpenAI's agent pipeline scanned the Linux kernel, reproduced 1,724 vulnerabilities and built 948 unique local root exploits, then wrote patches for maintainers. The full run cost about $1.2M in API credits.

OpenAI · 2026-11-04 · 3 sources checked

OpenAI social card for the research post A thousand fewer paths to root — OpenAI

OpenAI turned Linux kernel bug hunting into a pipeline: 2,308 findings in, 948 working root exploits and patches out.

Key specs

Unique root exploits	948
Vulnerabilities reproduced	1,724

Quick facts

Maker	OpenAI Cyber team
Target	Latest Linux kernel master, mid-October 2026
Pipeline	Discovery, validation, exploitability, patching, human review
Model	An upcoming OpenAI cyber model
End-to-end cost	$1,213,430 in API credits
Status after 10 days	61 fixed upstream, 43 staged, 299 under public discussion

What is it?

"A thousand fewer paths to root" is an OpenAI security research report on an agent pipeline that audits the Linux kernel at scale. Built on an upcoming OpenAI cyber model, it produced 2,308 candidate findings, reproduced 1,724 of them as real vulnerabilities and built 948 unique local privilege escalation exploits — bugs that let an unprivileged user become root. Human researchers reviewed the results before reporting them with candidate patches.

How does it work?

Scout agents first picked 5,034 audit targets reachable by low-privilege users, about 7% of the kernel's code. Analyst agents then scanned them, and triage agents wrote a proof of concept for each finding and ran it in a VM with the Kernel Address Sanitizer until it reproduced. Exploit-development agents tested the strongest candidates against a stock distro's protections (SMEP, SMAP, KASLR), and a final stage wrote patches. An early run showed why checks matter: 20 of 200 "successful" exploits reached root through a different zero-day than the one assigned.

Why does it matter?

The OpenAI report puts real prices on AI-scale offence and defence: a successful exploit cost $359 on average, and historical Pwn2Own and TyphoonPWN bounties would value the 948 exploits at 16x to 55x the run's cost. Its advice to defenders follows from that — rescan old code as models improve, revalidate against fast-moving upstream fixes, and keep systems patched. The agents also used 10 exploitation techniques that outside kernel experts had not seen, which OpenAI is not publishing.

Who is it for?

kernel maintainers, security teams and AppSec researchers

Frequently asked questions

How much did OpenAI's Linux kernel scan cost?: OpenAI reports the mid-October 2026 Linux kernel campaign cost $1,213,430 in API credits end to end. The largest stage was exploitation at $775,929; candidate discovery cost $316,549 and validation $73,330. A successful exploit averaged $359. OpenAI calls these figures an upper bound, since better harness engineering could cut the cost of each stage.
Have the Linux kernel vulnerabilities OpenAI found been fixed?: Partly. When OpenAI's scan and manual analysis finished after 10 days, 61 of the local privilege escalation bugs were already patched upstream, 43 had staging patches and 299 were being reported and discussed publicly. Following the Linux kernel community's advice, OpenAI started with critical vulnerabilities and is sending the rest gradually with Patch the Planet partners.
How does the new OpenAI cyber model compare with GPT-5.6 Sol on the kernel?: In a late-July 2026 scan with the same research harness, GPT-5.6 Sol found and exploited 166 unique Linux kernel privilege escalation bugs for $513,690 in API credits. The mid-October run with OpenAI's newer, unreleased cyber model found and exploited 948 for $1,213,430. OpenAI's takeaway is to rescan audited code whenever model capability improves.
Did OpenAI's pipeline work against Google's hardened kernelCTF target?: Yes. Run under Google's kernelCTF environment, a hardened kernel with a reduced attack surface, OpenAI's pipeline produced 25 information-leak proofs of concept and 21 local privilege escalation exploits over 112 hours for about $32,936 in API credits. OpenAI estimates those 21 would be worth about $1.5M in bounties if accepted, roughly 45x the cost.