OpenAI Limits GPT-5.6 Access After U.S. Request

The U.S. government imposed its first preemptive restriction on a frontier AI model, limiting access to OpenAI's GPT-5.6 family and Anthropic's Claude Mythos 5. OpenAI released three GPT-5.6 variants to 20 government-vetted companies, while Anthropic's Claude Mythos 5 was cleared for 100 institutions. The move sets a governance precedent for security research and enterprise integration, with OpenAI expecting broader access within weeks.

The U.S. government's first preemptive restriction on a frontier model release - applied simultaneously to OpenAI's GPT-5.6 family and Anthropic's Claude Mythos 5 - sets a governance precedent with direct implications for how security research, benchmarking, and enterprise integration of frontier models will work going forward. OpenAI released three models GPT-5.6 Sol, Terra, and Luna to roughly 20 government-vetted companies; Sol is the flagship, with OpenAI reporting 88.8% on Terminal-Bench 2.1 vs. Mythos 5 at 88.0% vendor-reported . Commerce Secretary Howard Lutnick separately cleared Anthropic to release Claude Mythos 5 to approximately 100 institutions. Government-gated previews concentrate early red-teaming inside approved channels, delay independent benchmarking, and shift first-impression narratives toward vendor-supplied data. OpenAI stated the vetting process should not become the "long-term default" and expects broader access within weeks.