OpenAI’s rogue AI model incident was worse than we thought

Two new reports offer nearly 130 pages of details on the OpenAI-Hugging Face cybersecurity incident, many of them previously unreleased.