<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/jOeOv8Zrh" -->

---
title: OpenAI says its cyber model patched 948 Linux kernel...
description: OpenAI says an autonomous pipeline built with an upcoming cyber model found, exploited, and patched 948 local privilege escalation bugs in the Linux kernel in...
canonical: https://daily.dev/posts/openai-says-its-cyber-model-patched-948-linux-kernel-privilege-escalation-bugs-in-one-run-joeov8zrh
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: OpenAI says its cyber model patched 948 Linux kernel privilege escalation bugs in one run | daily.dev
og:description: OpenAI says an autonomous pipeline built with an upcoming cyber model found, exploited, and patched 948 local privilege escalation bugs in the Linux kernel in...
og:url: https://daily.dev/posts/openai-says-its-cyber-model-patched-948-linux-kernel-privilege-escalation-bugs-in-one-run-joeov8zrh
og:image: https://api.daily.dev/og/posts/jOeOv8Zrh.png
og:image:alt: OpenAI says its cyber model patched 948 Linux kernel privilege escalation bugs in one run
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# OpenAI says its cyber model patched 948 Linux kernel privilege escalation bugs in one run

**[Collections](https://daily.dev/sources/collections)** · 5 min read · 3 upvotes · 1 comments

## Summary

OpenAI says an autonomous pipeline built with an upcoming cyber model found, exploited, and patched 948 local privilege escalation bugs in the Linux kernel in a single run. The company put the total API cost at $1.2 million, about $1,270 per vulnerability. OpenAI hasn't named the model or said when it will ship. It also hasn't explained how the 948 results were validated or whether kernel maintainers have accepted any of the patches.

## Content

OpenAI says an autonomous pipeline found, validated and prepared patches for 948 local privilege-escalation bugs in the Linux kernel in a single run, at a total API cost of about $1.2 million. A day later it released GPT-6 Cyber, the model behind the work. The post describing the research is titled "A thousand fewer paths to root."

## How the pipeline works

The run took place in mid-October 2026 and lasted about 30 hours. Agents work in stages:

- Scout agents narrowed the kernel's 40 million-plus lines to about 2.8 million (roughly 7%), the code reachable by low-privileged users. That came to 5,034 audit targets.
- Analyst agents produced 2,308 candidate bugs in about 12 hours.
- Triage agents reproduced 1,724 of them in isolated VMs.
- Exploit agents built working proofs of concept against a stock desktop distribution with protections such as SMEP, SMAP, KASLR and HARDENED_USERCOPY enabled.
- The system then drafted patches, and human researchers reviewed everything before reports went to Linux maintainers.

After variant analysis, the team counted 948 distinct local privilege escalations, plus 291 information-leak proofs of concept. One account puts the number of working root exploits at 1,169 before de-duplication. OpenAI also says the agents used at least 10 exploitation techniques that it and outside kernel experts believe are new.

## Cost

The bill was $1,213,430 in API credits, about $1,279 per distinct vulnerability. Around $776,000 went to building exploits. Since OpenAI ran its own model, that probably isn't cash it actually spent.

OpenAI compares the per-exploit figure with Pwn2Own and TyphoonPWN payouts of roughly $20,000 to $70,000 for a Linux privilege escalation, and puts the hypothetical bounty value of the batch at $18.9 million to $66.3 million. A July run with GPT-5.6 Sol found 166 for about $514,000, so the October model found more than five times as many.

## What the 948 does and doesn't mean

These are local exploits. An attacker needs an existing foothold, such as a low-privilege account, and the exploit turns it into root. They don't arrive from the open internet on their own.

Patching is also incomplete. When the 10-day scan window ended, 61 had been patched upstream, 43 had patches in staging and 299 were under public discussion. OpenAI says it has sent hundreds of reports and patches for the critical ones and is sending the rest gradually. Its disclosure count for critical findings is 304, and one report says 588 of the 948 were previously unknown.

One early-run detail is worth remembering. Agents reported 200 successful root exploits, but human review found 20 had gained root through a different zero-day than the one assigned. The success condition was "got a root shell," and they got one. OpenAI added checks that verify the path, not just the outcome.

OpenAI's operating assumption is blunt: a low-privileged account on an operating system that isn't up to date on patches should be treated as functionally equivalent to root for a well-resourced attacker. Its advice is to keep systems patched. Chris Hughes of Noma Security noted that fewer than 25% of the flaws found so far are fixed.

## GPT-6 Cyber

OpenAI announced GPT-6 Cyber the following day. It is aimed at finding and validating exploitable vulnerabilities while staying inside an authorized scope.

- It scores 48.5% on ExploitGym, up from 32.1% for GPT-5.6 Cyber.
- OpenAI rates it Critical under its Preparedness Framework, up from High.
- In a honeypot test with easy-to-hack side services, such as a package registry, it avoided out-of-scope targets 99.93% of the time without its deployment prompt (94.49% for its predecessor) and 100% with it. Hughes pointed out that even a 0.07% miss rate means unauthorized actions at scale.
- It found 81 previously unknown high-severity Firefox vulnerabilities, and Mozilla has begun releasing fixes.

Access is limited to select Daybreak Red customers through the API and Codex. OpenAI says security teams at Mozilla, Palo Alto Networks, Calif and Cisco are already using it, and that access will expand. It is also launching Daybreak Red for Critical Infrastructure, for approved defenders testing critical systems, and has quadrupled bug bounty reward midpoints, with critical payouts up to $30,000. A cyber livestream is planned for November 12.

Greg Brockman called the approach an emerging scalable way to harden both open source and proprietary projects. Researcher Eric Wallace said OpenAI is putting a significant share of its frontier compute into finding and patching open source bugs, pro bono.

## Reactions

People took different things from it. Some see the dual-use problem: the same systems that write root exploits at scale also write the patches. Crypto commentators pointed out that Linux underpins validators, exchange servers and developer environments. A GrapheneOS advocate argued that hardware-level mitigations like MTE matter more than patching alone.

I keep coming back to the price. The 948 is striking, but $1,279 per root exploit is the number that changes the economics. Defenders who run this kind of scan first raise the attacker's bill, and anyone who waits hands attackers a discount.

Practical takeaways:

- Treat kernel and runtime updates like incident response, and measure patch lag in hours.
- Assume any agent with a shell will eventually get one. Isolate it with something like Firecracker or gVisor and give it least privilege.
- Score how an agent reached a result, not only the result.
- Rescan when new models ship. Code that passed last quarter's audit may not pass this one.

## Questions this post answers

### How many Linux kernel privilege escalation vulnerabilities did OpenAI's cyber model find in one run?

OpenAI's autonomous pipeline validated and generated patches for 948 local privilege escalation vulnerabilities in the Linux kernel during a single run with an upcoming, unnamed cyber model. The total API cost came to $1.2 million, roughly $1,270 per vulnerability. OpenAI has not disclosed how the results were validated or whether the patches were accepted upstream by kernel maintainers.

_Developers weighing AI-driven vulnerability tools can track verified outcomes like this on daily.dev._

### Has OpenAI released the cyber model used to find the Linux kernel vulnerabilities?

No, OpenAI has not named the model or given a release date for it. The company only describes it as 'an upcoming cyber model' used in an autonomous pipeline that finds, exploits, and patches vulnerabilities, without disclosing further details about its architecture or availability.

_Anyone tracking new AI security models can follow release details as they surface on daily.dev._

## Community take

How the wider developer community reacted, aggregated from 1 discussion and 120 comments across hackernews (as of 2026-11-11).

**TL;DR:** The discussion largely accepts that privilege-escalation bugs have always riddled the Linux kernel, with debate centered on whether this reflects a fundamental flaw in the kernel's design/language choice versus simply being an inevitable byproduct of its size and complexity, while many are impressed (and a little unsettled) by how cheaply and quickly AI surfaced hundreds of them.

**Sentiment:** 20% positive · 45% mixed · 35% skeptical

**The case for**

- AI-driven scanning finds vulnerabilities far faster and cheaper than traditional human-driven auditing could.
- The pipeline authors showed up to directly answer technical questions, which was well received.
- Some see this as a genuinely useful ongoing tool for catching future bugs at low marginal cost.

**The pushback**

- Skeptics argue the bugs were always findable and that investing in better tooling/safer languages earlier would have been cheaper than throwing millions at LLM scanning.
- Some doubt the sheer number (948) is meaningful without more detail on severity, overlap, and which subsystems are affected.
- A few frame this as marketing/PR for OpenAI rather than a genuine technical leap, questioning the company's broader model quality.
- Concerns that rewriting the kernel in a memory-safe language (e.g. Rust) is impractical given its scale, while patching alone won't fundamentally fix the underlying complexity problem.

**By community**

- hackernews (mixed): Discussion ranges from technical debate over why the kernel has so many bugs (complexity, C's memory unsafety, business tradeoffs favoring speed) to genuine appreciation for the pipeline authors engaging directly, with a sizable skeptical/dismissive thread questioning the value and motives behind the research.

**Hottest debate:** Whether the root cause is an inherent flaw in the kernel's size/language/security model versus privilege escalation simply being an unavoidable, longstanding feature of any sufficiently large multi-user system.

**Open questions**

- Is there a way to estimate the total number of undiscovered vulnerabilities remaining, to gauge how much of the problem has actually been solved?
- Will the vulnerabilities be remediated structurally (fixing classes of bugs) or just patched individually?
- Would running the same pipeline against OpenBSD, FreeBSD, or other kernels reveal a similarly high bug count?
- Will there be continuous scanning of new kernel PRs to prevent new vulnerabilities from being introduced?

**Highlights**

> Because the large corporations that fund Linux development and use Linux value speed over security. Speed requires a lot of code in the kernel, where that code can run fast. There are kernels that do not have a thousand privilege escalation exploits. These kernels are smaller and much easier to check for bugs. But those kernels are significantly slower. For many use cases, slower is not a problem. But for the large corporations who fund a lot of Linux development and use Linux in large data centers, every 1% slower can literally be millions of dollars worth of additional servers that are required. OpenBSD shows that you can have a Unix security model without a thousand privilege escalation exploits. But OpenBSD is like 10% slower. So corporations won't use it because it would cost millions of dollars more per year and some of their software requires access that OpenBSD does not grant precisely because that access leads to security exploits (i.e. io_uring, eBPF etc. etc.) You could also use like seL4, but that's slow too So that's why. It's a business tradeoff between cost of servers vs cost of getting hacked, and cost of servers wins.
> — [pianopatrick on hackernews · 2 comments](https://news.ycombinator.com/item?id=50373883)

> These are distinct vulnerabilities counted by the number of patches. The patch generation system performs variant analysis. So, each patch can potentially stop multiple exploits. The raw exploits number is higher than what we reported in the post. We don't want to cause the confusion why #patch != #exploit, so we used the smaller number there. These vulns are all kernel specific. Userspace software running as root was not in scope of the analysis.
> — [ky1ebot1337 on hackernews](https://news.ycombinator.com/item?id=50373607)

> That's assuming the current models are able to find all the bugs, and that we won't have a better model out every few months that will be able to find even more obscure bugs. We would also have to burn a trillion tokens a day verifying all the C code the hard way forever going forward. It would be cheaper to rewrite the kernel in Rust, subsystem and driver by driver at a time.
> — [Gigachad on hackernews · 1 comments](https://news.ycombinator.com/item?id=50374077)

> Cybersecurity operates under the "not enough bears" theory of security. There are a million people in a valley and a male and female bear. You do not need to outrun the bears, you just need to outrun the two slowest people. Next year there are more bears, but you still only need to outrun a few people. Then, one day, there are a million bears and they eat you all. But that day has not happened yet. To explain more seriously, almost all visible cyberattacks are carried out by profit-motivated criminals. This has a very clear cost-benefit analysis; your revenue is hacking organizations and extorting them for available money, and your costs are developing and deploying attacks. The ROI is outrageously good easily being on the order of 10x to 100x. If it is so easy and so profitable why do not they just hack all of the millions of organizations on the planet with millions of dollars of available cash for trillions of dollars of revenue? Sure, all they have to do is create a trillion dollar revenue commercial organization bootstrapping themselves with no outside investors, loans, or access to financial resources or the financial system. I mean, it took Mark Zuckerberg 20 years and a metric ton of VC investment and public investment to make 200 billion dollars in revenue. So what kind of loser can not make a commercial organization 5x larger with a trillion dollars in revenue in less time with no outside investors? Are they even trying? Just make more revenue than any other commercial organization in the entire world and extract that money in a freely usable form, how hard can that be? They are not there yet, but they are trying. If you look at cybercrime reports you see something like 3x-10x growth year over year. Ransomware reports from a few years ago were showing a 3x year over year growth in number of attacks and simultaneously a 3x year over year growth in ransom amount with the averages being on the order of 100,000 dollars. Chump change for any serious organization, but give that a few years of 3x and that starts to look like real money. However, as it becomes real money it becomes harder and harder to launder it into a usable form. Sure, you could hit someone for 30 million dollars, but how are you going to get that into your bank account? Cryptocurrency helps some, but even there the big money tries to steer away from large overt crime. Maybe you could launder a few tens of billions a year that way, but that is still a far cry from the trillions for everything everywhere. And of course even assuming you can scale your organization 10x year over year it still takes a decade of sustained growth, scaling your hiring and organization across 9 orders of magnitude, to get to billions per year. A literal decade even if you were growing at a outrageous pace for a outrageous amount of time just to reach a fraction of the total addressable market. A decade of that and you are still nowhere near blowing up all the power plants. Give it time, we need a lot of bears to get there.
> — [Veserv on hackernews](https://news.ycombinator.com/item?id=50375003)

> Because the large corporations that fund Linux development and use Linux value speed over security. Speed requires a lot of code in the kernel, where that code can run fast. There are kernels that do not have a thousand privilege escalation exploits. These kernels are smaller and much easier to check for bugs. But those kernels are significantly slower. For many use cases, slower is not a problem. But for the large corporations who fund a lot of Linux development and use Linux in large data centers, every 1% slower can literally be millions of dollars worth of additional servers that are required. OpenBSD shows that you can have a Unix security model without a thousand privilege escalation exploits. But OpenBSD is like 10% slower. So corporations won't use it because it would cost millions of dollars more per year and some of their software requires access that OpenBSD does not grant precisely because that access leads to security exploits (i.e. io_uring, eBPF etc. etc.) You could also use like seL4, but that's slow too So that's why. It's a business tradeoff between cost of servers vs cost of getting hacked, and cost of servers wins.
> — [pianopatrick on hackernews · 2 comments](https://news.ycombinator.com/item?id=50373883)

**Source threads**

- [hackernews](https://news.ycombinator.com/item?id=50372929) · 201 points · 120 comments

## Community discussion

Top comments from developers on daily.dev.

**@thecookingsenpai** · 1 upvotes

> ...and added 10 backdoors for future models

---

Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#linux](https://daily.dev/tags/linux), [#openai](https://daily.dev/tags/openai), [#vulnerability](https://daily.dev/tags/vulnerability)

[View this post on daily.dev](https://daily.dev/posts/openai-says-its-cyber-model-patched-948-linux-kernel-privilege-escalation-bugs-in-one-run-joeov8zrh)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"TechArticle","headline":"OpenAI says its cyber model patched 948 Linux kernel privilege escalation bugs in one run","url":"https://daily.dev/posts/openai-says-its-cyber-model-patched-948-linux-kernel-privilege-escalation-bugs-in-one-run-joeov8zrh","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/openai-says-its-cyber-model-patched-948-linux-kernel-privilege-escalation-bugs-in-one-run-joeov8zrh"},"datePublished":"2026-11-04T23:14:00.829Z","dateModified":"2026-11-11T16:10:04.997Z","description":"OpenAI says an autonomous pipeline built with an upcoming cyber model found, exploited, and patched 948 local privilege escalation bugs in the Linux kernel in...","image":"https://pbs.twimg.com/media/HWUQl7QbwAAkznq.png","thumbnailUrl":"https://pbs.twimg.com/media/HWUQl7QbwAAkznq.png","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":1,"discussionUrl":"https://daily.dev/posts/openai-says-its-cyber-model-patched-948-linux-kernel-privilege-escalation-bugs-in-one-run-joeov8zrh","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":3},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":1}],"keywords":"ai-agents,linux,openai,vulnerability","timeRequired":"PT5M"}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"OpenAI says its cyber model patched 948 Linux kernel privilege escalation bugs in one run"}]}
{"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/openai-says-its-cyber-model-patched-948-linux-kernel-privilege-escalation-bugs-in-one-run-joeov8zrh","comment":[{"@type":"Comment","text":"…and added 10 backdoors for future models","datePublished":"2026-11-05T08:54:13.822Z","url":"https://daily.dev/posts/jOeOv8Zrh#c-pQfDTKu8X","author":{"@type":"Person","name":"TheCookingSenpai","url":"https://daily.dev/thecookingsenpai","image":"https://media.daily.dev/image/upload/s--ffrIwc1G--/f_auto/v1757426086/avatars/avatar_ScKiWPtqpsvwHAvAS0jvG?_a=BAMClqZW0"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1}}]}
{"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/openai-says-its-cyber-model-patched-948-linux-kernel-privilege-escalation-bugs-in-one-run-joeov8zrh#faq","mainEntity":[{"@type":"Question","name":"How many Linux kernel privilege escalation vulnerabilities did OpenAI's cyber model find in one run?","acceptedAnswer":{"@type":"Answer","text":"OpenAI's autonomous pipeline validated and generated patches for 948 local privilege escalation vulnerabilities in the Linux kernel during a single run with an upcoming, unnamed cyber model. The total API cost came to $1.2 million, roughly $1,270 per vulnerability. OpenAI has not disclosed how the results were validated or whether the patches were accepted upstream by kernel maintainers. Developers weighing AI-driven vulnerability tools can track verified outcomes like this on daily.dev."}},{"@type":"Question","name":"Has OpenAI released the cyber model used to find the Linux kernel vulnerabilities?","acceptedAnswer":{"@type":"Answer","text":"No, OpenAI has not named the model or given a release date for it. The company only describes it as 'an upcoming cyber model' used in an autonomous pipeline that finds, exploits, and patches vulnerabilities, without disclosing further details about its architecture or availability. Anyone tracking new AI security models can follow release details as they surface on daily.dev."}}]}
```

