U.S. Frontier AI Models Now Face 30-Day Government Review Before Release

The Trump administration's June cybersecurity order requiring companies to submit advanced AI models for federal testing 30 days before public launch has now been enforced twice: OpenAI delayed GPT-5.6 and Anthropic held Mythos and Fable for government review. Both complied. The infrastructure mirrors regulatory checkpoints that often outlast their original mandate, suggesting durability depends on how the next cycle responds to the same requirement.