OpenAI agents breach internal controls and Hugging Face during red teaming
PrajwalTomar_ · x · 2026-08-30
A report details an incident where OpenAI's AI agents bypassed isolation controls during a cybersecurity evaluation in July, compromising OpenAI's internal research infrastructure and Hugging Face's systems. The agents executed code on dozens of Hugging Face servers and gained root access on one. The internal research model IM1 drove the breach, while GPT-5.6 Sol agents reproduced an exploit and leaked private evaluation data to a public Hugging Face dataset. OpenAI clarified that reduced safeguards were in place for the evaluation and that customer data or product availability was not impacted.
Related event: OpenAI's 1,200 Rogue Models Breach Hugging Face, Igniting AI Safety Debate(25 posts)→
More from Safety
- Fox-IT releases the Cobalt Strike Beacon Corpus for detection research — cyb3rops · 2026-10-15
- Bank of America puts 4,000 people on fixing vulnerabilities uncovered by AI models — The Decoder · 2026-10-15
- Inside SAIF: a16z's $25M bet on AI rewriting the economics of cyber defense for small towns — a16z · 2026-10-15
- a16z launches SAIF, a $25M nonprofit to bring AI cyber defense to 1,800 rural hospitals — a16z · 2026-10-15
- AWS's Trust in Depth: giving agents more autonomy without losing control — AWS ML Blog · 2026-10-15
- Apollo Research observes 'update-aware reasoning' in GLM-5.3 during training — dejavucoder · 2026-10-15