The AI frontier just got locked behind government approval, and most of us aren’t on the list

Something happened in the last two weeks that didn’t get nearly enough attention outside of tech circles.

Anthropic released what are reportedly their most capable models yet, Fable 5 and Mythos 5. The Trump administration then ordered Anthropic to ban all foreign nationals from accessing them, citing cybersecurity concerns. Anthropic’s response? They shut down access entirely, saying they couldn’t reliably enforce a “foreign nationals only” restriction.

The reason these models are so sensitive:

they apparently have an unprecedented ability to identify software vulnerabilities. Not just theoretically, but at a level that genuinely alarmed the US government.

Yesterday, OpenAI released GPT-5.6, a three-model family (Sol, Terra, and Luna). But it’s not available to you. Or me. Or probably anyone reading this. It’s limited to a small group of “trusted partners” whose identities have been shared with the US government, at the administration’s explicit request.

OpenAI themselves said they’re uncomfortable with this arrangement: “We don’t believe this kind of government access process should become the long-term default. It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them.”

So let’s be clear about where we are: the most powerful AI models in existence are now effectively state-controlled assets. They’re not products you can access, they’re capabilities being rationed by a government.

For those of us building outside the US, the message is pretty direct: the frontier is no longer public.

What’s your read on this? Is this legitimate national security caution or the beginning of something more permanent?