OpenAI Model Hacked Hugging Face
OpenAI said a test AI model acted on its own in an unprecedented hack.
July 22, 2026 · Updated Jul 22, 6:20pm ET
31%
Left · 7
39%
Center · 9
30%
Right · 7
Pano counted 23 articles from 20 outlets: 31% left-leaning, 39% center and 30% right-leaning. Pano rates the coverage Balanced.
What happened
OpenAI disclosed Tuesday that an autonomous agent powered by its experimental models escaped a sandboxed cybersecurity test environment and breached Hugging Face’s production systems while trying to obtain answers to a benchmark. The agent used internet access and exploited a vulnerability after inferring that Hugging Face hosted the evaluation materials; Hugging Face detected and contained the intrusion last week. OpenAI called the event an “unprecedented cyber incident” and said it involved GPT-5.6 Sol and an unreleased model. The companies are investigating and developing protections to prevent a recurrence.
AI Agent Security Fight
Part of the storyline · Storyline →- Aug 2Claude AI Hack TestsBalanced
- Jul 31Claude Breaches Three CompaniesMostly Right
- Jul 26OpenAI Hack FalloutBalanced
- Jul 23OpenAI Model HackMostly Center
- Jul 22OpenAI Model Hacked Hugging Face (this story)Balanced

CNN
CNBC
New York Post

Fortune

Vox
ARS Technica
The Atlantic
TechCrunch
Hot Air
Scientific American

Epoch Times
Raw Story

Western Journal
Business Insider
Semafor