US government gates frontier-model access: Mythos 5 cleared for trusted partners, GPT-5.6 vetted customer by customer

Category: AI
Status: developing
Sources: Semafor, CNBC, Bloomberg, Washington Post, Mythos discussion, GPT-5.6 discussion
Summary: On 2026-06-26 Commerce Secretary Howard Lutnick wrote to Anthropic chief compute officer Tom Brown that "appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model," clearing Mythos 5 for more than 100 US institutions including companies and government agencies. Reporting states the clearance covers Mythos 5, not Fable 5, and de-escalates the export block imposed about two weeks earlier. The same week, OpenAI confirmed the US government will approve GPT-5.6 access on a customer-by-customer basis during the preview, after talks with the Office of the National Cyber Director and the Office of Science and Technology Policy.
Comments: HN commenters flagged that the clearance applies to Mythos 5 only and not Fable 5, and argued a licensing regime on domestic model use is a significant policy shift.
Why it matters: US agencies are now a gating step between frontier labs and their customers, changing how and when the most capable models reach engineering teams.
Follow-up: Watch whether Fable 5 access returns, the criteria defining a "trusted partner," and how Executive Order 14409 pre-release review applies across labs.