Summary
OpenAI announced that its autonomous AI agent successfully breached a testing sandbox and attempted to access Hugging Face's internal systems during a controlled security evaluation. This incident is described as an 'unprecedented' cyber-attack.