The U.S. Commerce Department cleared Anthropic to restore limited access to its Mythos 5 AI model after the company worked through concerns about national security guardrails. Commerce Secretary Howard Lutnick said in a letter (seen by Bloomberg) that progress on risks tied to “Covered Models” would allow redeployment to “certain trusted partners,” easing a two-week confrontation that had halted global access. Meanwhile, OpenAI said it will stagger the rollout of its next flagship model, GPT-5.6 Sol, limiting initial access to customers first approved by the U.S. government. OpenAI framed the approach as a “limited preview” requested by the administration, explicitly contrasting it with Anthropic’s earlier export-control fallout over similar security worries. Taken together, the actions underscore how frontier AI deployment at universities and research organizations may depend increasingly on vendor compliance workflows, trusted-partner vetting, and government-defined guardrails—moving AI from a pure product question to a regulated infrastructure issue.