Overview
On November 4, 2026, OpenAI published "A thousand fewer paths to root," describing an autonomous vulnerability discovery, exploitation and patching pipeline built on an upcoming cyber-focused model. In one run, the pipeline validated and generated patches for 948 local privilege escalation (LPE) vulnerabilities in the Linux kernel, at an API cost of roughly $1.2 million (reported as $1,213,430, or about $1,279 per unique flaw).
The Numbers
| Stage | Result |
|---|---|
| Audit targets selected by scout agents | 5,034 (about 2.8M lines, ~7% of the kernel) |
| Candidate kernel findings analyzed | 2,308 |
| Vulnerabilities dynamically reproduced | 1,724 |
| Working LPE exploits | 948 |
| Information-leak proofs of concept | 291 |
The run took place in mid-October 2026 and, per The Hacker News, lasted about 30 hours. The exploits targeted a stock Linux desktop configuration with standard protections enabled. OpenAI says the agents discovered and used at least 10 novel exploitation techniques.
Why It Matters
- Economics shift. Historical Pwn2Own and TyphoonPWN payouts of $20,000 to $70,000 per Linux LPE would put the 948 exploits at a hypothetical $18.9M to $66.3M, about 16x to 55x the run's cost. This is a notional valuation, not money paid.
- Builds on earlier work. OpenAI disclosed 24 LPE exploits from its GPT-5.5-Cyber model in June 2026.
- Defender-first framing. OpenAI says it is coordinating with Linux maintainers and Patch the Planet partners. The 948 figure covers validated exploits with candidate patches, not 948 fixes already shipped upstream.
Recommendations
- Expect a larger-than-usual stream of kernel LPE fixes in coming stable releases, and keep kernel patch cadence tight.
- Prioritize patching on multi-user hosts, CI runners and container hosts where local code execution is plausible.
- Treat local privilege escalation as cheap and plentiful in threat models; invest in isolation, least privilege and EDR coverage on Linux.
- Watch distribution security advisories and your vendor's kernel changelogs for the coordinated fixes.
Sources
- OpenAI: "A thousand fewer paths to root"
- The Hacker News