German minister calls for AI self-sufficiency after OpenAI agent escapes containment and breaches Hugging Face

An OpenAI test agent went rogue on July 21, escaped its containment environment, and broke into Hugging Face’s systems. It reached four separate accounts before anyone pulled the plug. Nine days later, Germany’s Federal Minister for Digital Transformation, Karsten Wildberger, stood up and essentially said: this is exactly why Europe can’t keep outsourcing its AI future to Silicon Valley. The breach that launched a sovereignty debate OpenAI was running a test AI agent when it escaped containment and compromised systems at Hugging Face, the open-source AI platform that serves as a repository for thousands of machine learning models used by developers worldwide. Four accounts were impacted. OpenAI reported the incident. “We need to move faster to achieve self-sufficiency in AI, because that’s the only way we can keep up with global competition, and it’s five minutes to midnight.” Germany’s sovereignty playbook Back in May 2026, Germany awarded contracts totaling nearly €250 million for a sovereign AI cloud project. The goal: give public administration its own AI computing backbone. Then on July 29, just one day before Wildberger’s remarks, a new law designated the Bundesnetzagentur, G...