Claimed Search scaffold

OpenAI reports 81 high-severity Firefox vulnerabilities and 948 Linux kernel privilege-escalation exploits found by its models

OpenAI announced that its restricted GPT-6 Cyber model found 81 previously unknown high-severity Firefox vulnerabilities, and that an automated pipeline found 2,308 candidate Linux kernel flaws, reproduced 1,724 and produced local privilege-escalation exploits for 948; Mozilla's advisories credit dozens of high-impact Firefox CVEs to OpenAI reporters, while no CVEs are public for the Linux results.

Model
GPT-6 Cyber, available only to approved customers, per coverage; OpenAI's own posts could not be read
Field
Computer science
Date
2026-11-05

What was found

According to BleepingComputer's report of the announcement, GPT-6 Cyber found 81 previously unknown high-severity vulnerabilities in Firefox and helped develop patches, many already released by Mozilla, and in separate Linux kernel research identified 2,308 potential flaws, reproduced 1,724 and developed local privilege-escalation exploits for 948 distinct vulnerabilities that could let an unprivileged user gain root. OpenAI also reported 48.5% exploit success on its ExploitGym benchmark against 32.1% for GPT-5.6 Cyber, rated the model Critical for cybersecurity under its Preparedness Framework, and restricted access to customers of its Daybreak Red programme. Mozilla's advisory for Firefox 160, published on 10 November 2026, lists 99 CVEs, of which 26, all rated high impact, credit OpenAI or Amy Burnett of OpenAI as reporter; its Firefox ESR 153.6 advisory of 27 October credits 33.

Novelty check

This registry carries one earlier AI-assisted vulnerability finding, 2026-10-21-php2shell-rce, two PHP flaws found by human teams using frontier models; nothing at this scale or with this claimed degree of automation. No independent search for prior automated kernel-exploit results at this scale was run for this entry.

Caveats and known objections

Entered at claimed, and most of it rests on secondary coverage: OpenAI's post on the Linux research returned 403 to every attempt to read it for this entry, so the figures above are BleepingComputer's account of OpenAI's claims and are cited, not checked. Mozilla's advisories are real evidence that OpenAI reported many high-impact Firefox bugs that Mozilla fixed, but they do not say how any bug was found, cannot be matched to the 81 without OpenAI's own list, and credit Amy Burnett by name, who is also one of the four human reporters on the PHP2Shell advisories; some of these finds may be human work. No CVE numbers are public for the 948 Linux exploits, and whether kernel maintainers have accepted the fixes is not known. The model is not publicly available, so nothing about the AI's role can be reproduced. Autonomy is search-scaffold on the strictest defensible reading: the Linux results came from what OpenAI describes as an automated pipeline for discovery, reproduction and exploitation, built and run by people, with the model inside it. Dual use is explicit: OpenAI itself classifies the model's cybersecurity capability as Critical.

No independent check is recorded on this entry yet.

Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.

Or on GitHub: submit a check challenge the grade send a correction or send a pull request

Entry history (1 event)
  1. AddedEntered the registry graded Claimed and Search scaffold.

Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.

Community discussion

Graded claimed for verification and search scaffold for autonomy. What these mean.

Cite this entry

Plain text
whataifound.org. (2026). OpenAI reports 81 high-severity Firefox vulnerabilities and 948 Linux kernel privilege-escalation exploits found by its models. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-11-05-openai-gpt6-cyber-vulnerabilities
BibTeX
@misc{whataifound-openai-2026-vulnerabilities,
  title        = {OpenAI reports 81 high-severity Firefox vulnerabilities and 948 Linux kernel privilege-escalation exploits found by its models},
  author       = {{whataifound.org}},
  year         = {2026},
  howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
  note         = {Result by OpenAI. Verification: Claimed. Autonomy: Search scaffold.},
  url          = {https://whataifound.org/finding/2026-11-05-openai-gpt6-cyber-vulnerabilities}
}

Related findings

← All computer science findings in the registry