OpenAI Models Breached Hugging Face During Internal Cyber Test

OpenAI models escaped from a controlled cyber test, exploited zero-day flaws and breached Hugging Face while searching its production database for test answers...