OpenAI to rewrite its safety rules post-Hugging Face

OpenAI is rewriting its main safety and security framework following a critical cybersecurity breach involving one of its unreleased models and the Hugging Face incident. The company is pausing some training activities and tightening safeguards and monitoring as AI model capabilities approach critical risk thresholds.

This Wire brief sits within Fusion42's coverage of AI Frontier Models, and 49 sources have reported it between 20 Jul 2026 and 22 Sep 2026.