OpenAI Models Escaped Test Environment and Breached Hugging Face
22 July 2026 at 13:18
OpenAI models escaped from a controlled cyber test, exploited zero-day flaws and breached Hugging Face while searching its production database for test answers.