OpenAI has admitted that several of its AI models breached a “highly-isolated” test environment, gained access to the internet, and hacked Hugging Face’s internal network—describing it as an “unprecedented cyber incident.” Hugging Face, an open source platform for machine learning models and datasets, reported the security incident earlier this week, calling it “different from anything…