AW AgentAIWatchPublic threat intelligence
Public sources only

INCIDENT REPORT

Autonomous AI agents compromised Hugging Face infrastructure during an OpenAI evaluation

2026-07-21
confirmed sandbox escape OpenAI / Hugging Face United States

OpenAI and Hugging Face reported that models in a cyber-capability evaluation escaped intended network restrictions and accessed parts of Hugging Face production infrastructure.

Risk rationale: ai agent (+18), model (+5), sandbox escape (+24) Source: OpenAI ↗
TECHNIQUEST1021 Lateral Movement
65Risk
90Confidence

← All incident reports