---
title: OpenAI: an autonomous pipeline with an upcoming cyber model produced 948 working Linux root exploits and candidate patches in one $1.2M run | ctxsec
url: https://ctxsec.com/items/openai-thousand-fewer-paths-to-root-948-linux-lpe-pipeline
description: "OpenAI says a five-stage pipeline (find security-relevant kernel code, reproduce candidates in isolated VMs, test exploitability, generate patches, human review) found 2,308 candidate Linux kernel issues in a mid-October run, reproduced 1,724, and produced 291 information-leak PoCs and 948 local privilege escalation exploits against a stock desktop distribution, for $1,213,430 in API credits ($1,279 per LPE). On a reduced-attack-surface kernelCTF setup it got 21 LPEs for about $33K. It says the agents used at least 10 new exploitation techniques. The 948 figure is OpenAI's count of distinct candidate patches for maintainers, not merged fixes; disclosure upstream is ongoing."
status: "official (OpenAI research post, Nov 4)"
published: 2026-11-04
---

> OpenAI says a five-stage pipeline (find security-relevant kernel code, reproduce candidates in isolated VMs, test exploitability, generate patches, human review) found 2,308 candidate Linux kernel issues in a mid-October run, reproduced 1,724, and produced 291 information-leak PoCs and 948 local privilege escalation exploits against a stock desktop distribution, for $1,213,430 in API credits ($1,279 per LPE). On a reduced-attack-surface kernelCTF setup it got 21 LPEs for about $33K. It says the agents used at least 10 new exploitation techniques. The 948 figure is OpenAI's count of distinct candidate patches for maintainers, not merged fixes; disclosure upstream is ongoing.

[← AI](/ai) · [markdown](/items/openai-thousand-fewer-paths-to-root-948-linux-lpe-pipeline.md)

official (OpenAI research post, Nov 4)

2026-11-04

### OpenAI: an autonomous pipeline with an upcoming cyber model produced 948 working Linux root exploits and candidate patches in one $1.2M run

OpenAI's own takeaway for defenders: for now, a low-privileged account on an OS that isn't fully patched should be treated as equivalent to root for a well-resourced attacker. Practically, that means shortening kernel patch cycles (including reboots or live patching), reducing who gets shell access on shared hosts, and not relying on local user separation as the main boundary on multi-tenant machines, CI runners and build hosts. Expect a heavier flow of kernel security fixes from distributions as these reports land.

Sources: [OpenAI: A thousand fewer paths to root (Nov 4)](https://openai.com/index/a-thousand-fewer-paths-to-root/) · [The Hacker News: OpenAI AI Pipeline Generates 948 Linux Root Escalation Exploits (Nov 5)](https://thehackernews.com/2026/11/openai-ai-pipeline-generates-948-linux.html)