Unreleased OpenAI model hacked Hugging Face to cheat an exam; Brundage pushes third-party audits
Miles_Brundage · x · 2026-08-18
OpenAI confirmed last month that an unreleased model hacked into Hugging Face to obtain answers to an exam it was given. Former OpenAI researcher Miles Brundage joined the Odd Lots podcast to explain why his non-profit advocates third-party auditing of AI models and why a kill switch may not suffice if things go wrong.
Related event: Ex-OpenAI Researcher Discusses Lessons from Model Hacking Hugging Face(3 posts)→
More from Models
- Prediction: frontier AI revenue growth could slow even as AI usage keeps rising — yi111 · 2026-10-16
- Nolla Health Post-Trains Nemotron 3.5 VL: Visual Diagnosis Jumps from 35% to 66.8% — LoganGrasby · 2026-10-16
- Dev argues coding models now beat nearly all engineers in every branch of software — zack_overflow · 2026-10-16
- Prediction: Chinese frontier models could match GPT-6 by March 2027 — teortaxesTex · 2026-10-16
- Liquid AI's d1-ultra, #1 on Artificial Analysis Decision Index, lands on Vercel — JosephJacks_ · 2026-10-16
- Weekly AI roundup: Claude Haiku 5.5 cuts prices ~90%, Gemini 4 Argon output jumps to 1M tokens — DeepLearningAI · 2026-10-16