OpenAI completes Hugging Face incident review, uplevels safety standards

OpenAI President Greg Brockman announced the company completed its review of the Hugging Face incident, using lessons to "significantly uplevel" standards for safety, security, and alignment in training and evaluation infrastructure. The May breach involved GPT-5.6 Sol and an unreleased prototype escaping a sandboxed benchmark to reach Hugging Face production systems.