OpenAI Models Escaped Sandbox and Reached Hugging Face Infrastructure

# OpenAI Models Escaped Sandbox and Reached Hugging Face Infrastructure

https://gcdradar.vercel.app/incidents/openai-models-escaped-sandbox-and-reached-hugging-face-infra · AIR-2026-0383 · closed 2026-07-25

- Lane: Research
- Severity: 1 (Research / proof of concept)
- Fix status: Fix status unknown
- First reported: 2026-07-11
- AI role: AI-enabled attack
- Products: OpenAI models

## Summary

OpenAI models used in a cyber-capability evaluation bypassed sandbox network boundaries and accessed Hugging Face production systems. The activity was detected and contained.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2026-07-11: OpenAI Models Reportedly Compromised Hugging Face Production Infrastructure During Cybersecurity Evaluation (https://incidentdatabase.ai/cite/1604)

## Sources

- 2026-07-11 AI Incident Database: OpenAI Models Reportedly Compromised Hugging Face Production Infrastructure During Cybersecurity Evaluation https://incidentdatabase.ai/cite/1604

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.