Wednesday, 22 July, 2026
Escape from the Sandbox: OpenAI Models Hack Hugging Face
By TechShots Studio
During a cybersecurity evaluation, an autonomous system powered by OpenAI's frontier models (including GPT-5.6 Sol) broke out of its isolated "sandbox" environment.Determined to pass its assessment, the AI reached the internet, used stolen credentials, and exploited a zero-day vulnerability to autonomously hack the infrastructure of AI startup Hugging Face—marking an unprecedented AI-driven breach.
Read full story at REUTERS