OpenAI's run reproduced 1,724 Linux bugs, with 948 working privilege escalation exploits
OpenAI · x · 2026-11-05
The run's full funnel: 2,308 candidate findings led to 1,724 reproduced vulnerabilities, 291 working information-leak PoCs, and 948 working local privilege escalation exploits—all effective on a stock Linux desktop configuration with standard security protections enabled.
More from Safety
- Palisade defends AI self-replication proof-of-concept, says wild spread is inevitable — JeffLadish · 2026-11-05
- Guidelight ships Capability Testing standard v1.0 with sandbagging-detection practice — dfrsrchtwts · 2026-11-05
- MCP over HTTP: validating OAuth tokens isn't enough—check the audience claim — alvelda · 2026-11-05
- keltan updates Astra sandbagging analysis: no large updates from new evidence — repligate · 2026-11-05
- Texas lawmaker wants data centers to pay their own energy, recycle water, get community consent — Scobleizer · 2026-11-05
- Residential proxy exits detected on White House network: 236 exits via 28 services — matthew_d_green · 2026-11-05