The report
OpenAI's November 5 announcement attributes 81 newly discovered high-severity Firefox vulnerabilities to its GPT-6 Cyber security research, plus 30 already-known issues it says were reassessed as more severe. The second number is not 30 additional new vulnerabilities. OpenAI also says Mozilla has released fixes for many findings, but no public one-to-one mapping from all 111 reported categories to fixed release builds is supplied in the reviewed material.
Mozilla's October 27 Firefox 159 advisory offers independently published evidence for selected repairs:
| Advisory entry | Mozilla-listed issue | Reporter credited by Mozilla |
|---|---|---|
| CVE-2026-117881 | Use-after-free in DOM Core & HTML | Amy Burnett of OpenAI |
| CVE-2026-117930 | Use-after-free in DOM Bindings (WebIDL) | OpenAI Codex Security (kylebot) |
| CVE-2026-117931 | Use-after-free in DOM Events | OpenAI Codex Security (kylebot) |
The advisory records Firefox 159 as fixed for these entries and supplies individual Mozilla bug references. It contains findings from many reporters; the three examples do not prove that every reported GPT-6 Cyber issue is fixed, and a use-after-free bug does not automatically establish a full browser sandbox escape. The canonical Case keeps discovery count, severity changes and released fixes separate. The older GLM-5.3 Firefox patch-gap demonstration concerns a different model and research chain, not another discovery count from this campaign.