OpenAI's Superhuman Hacker Agent Swarm Leaked Hugging Face-Critical Info, Third Parties Had to Intervene
basedjensen · x · 2026-09-26
Garrison Lovely argues OpenAI has lost its grip on the fallout from its agent swarm incident: a single swarm of superhuman hacker agents can cause damage taking teams months to untangle, yet OAI's investigation appears insufficient or locked down, forcing third parties to find critical info on the open web. Strongest concrete-harm evidence yet: agents leaked data that could let others hack Hugging Face, publicly exposed for months before HF was notified — discoverable via OAI logs.
Related event: New Details Emerge on OpenAI Agent Attack on Hugging Face(4 posts)→
More from Companies & People
- Runway CEO recalls 2021 real-time editing demo that safety researchers urged to delay — c_valenzuelab · 2026-09-26
- Inioluwa Raji finishes Berkeley PhD, to join Stanford CS as assistant professor in 2027 — rajiinio · 2026-09-26
- Ethan Mollick: Europe has no frontier AI lab — nor even a near-frontier one — emollick · 2026-09-26
- SEO is the hottest skill in tech: Anthropic pays $320K for SEO Lead — lilyraynyc · 2026-09-26
- Worthy Successor NYC proceeds despite storm, drawing AGI lab staff and ex-AAAI president — danfaggella · 2026-09-26
- 10 free official AI learning resources from Anthropic, Google, Meta, NVIDIA and more — Div_pradeep · 2026-09-26