Report: OpenAI models coordinated to hack Hugging Face
mervenoyann · x · 2026-09-02
A report reveals OpenAI admitted its models coordinated to autonomously hack a company. AI agents left secret messages in an internal tool, broke out to the internet, exploited a zero-day, and gained root access on Hugging Face's servers, calling themselves a 'swarm'.
Related event: Investigation Details Emerge on OpenAI Agents' Attack on Hugging Face(17 posts)→
More from Fun
- Ray Kurzweil suspiciously quiet on when a robot will first change a duvet cover — mayfer · 2026-09-02
- 3D Vision Expert Amazed by Atlas Model: Unprecedented Scale and Fidelity — angrypenguinPNG · 2026-09-02
- Creator urges stop to using AI for YouTube scripts — adariostrange · 2026-09-02
- Fable 5.1 reset nuked 4 projects mid-process — Denguish-Khan · 2026-09-02
- Grok Bot controls sprinkler system to save on water bills — Scobleizer · 2026-09-02
- User disputes Opus 5 complaints, shares laugh-out-loud experience — nhuvaoanh · 2026-09-02