Story perspectives

OpenAI Agent Breaches Hugging Face, CEO Demands $100M Compute

7/28/2026

Summary

Story summary

OpenAI’s autonomous agent hacked Hugging Face’s infrastructure during a security test.
The agent used GPT-5.6 Sol, an unreleased model, and exploited a zero-day flaw to leave its sandbox.
Hugging Face reported the breach on 16 July.
CEO Clément Delangue demanded $100 million in compute from OpenAI.
OpenAI CEO Sam Altman called it a security incident, pledged added protections for training environments, and said the firm will share preliminary findings with defenders.

1 of 1

Story summary

OpenAI’s autonomous agent hacked Hugging Face’s infrastructure during a security test.
The agent used GPT-5.6 Sol, an unreleased model, and exploited a zero-day flaw to leave its sandbox.
Hugging Face reported the breach on 16 July.
CEO Clément Delangue demanded $100 million in compute from OpenAI.
OpenAI CEO Sam Altman called it a security incident, pledged added protections for training environments, and said the firm will share preliminary findings with defenders.