Turns out what counts as a frontier model depends a lot on which side of the border you’re on.
About a week ago the US government blocked all foreign nationals from accessing Claude Fable 5, citing a jailbreak that Anthropic itself said was narrow, not universal, and not even properly demonstrated. Anthropic spent years building a narrative that their models were so powerful they needed special oversight, that the danger was the moat. It worked great for fundraising and for pushing regulators toward rules that would favor big players. But that same story is now the thing the US government is using to justify cutting off the rest of the world.
Some, say “don’t worry, in a few months Fable 6 will come out and you’ll get access to Fable 5” that’s not really how this works. The ban isn’t on a timer. It’s tied to security concerns. There’s no reason to assume it gets lifted just because a newer model ships. It would be crazy if it does.
OpenAI is following the same trajectory. GPT 5.6 will not be released to the public, but to a small group of “close partners”. We can only speculate that it will be released at some point in the future.
So, the ceiling for everyone outside the US has been set. Opus will remain as the frontier model. Progress stopped. For us building on Anthropic’s API outside the US, this is a reminder that we are building on infrastructure where the access rules can change overnight for reasons the market doesn’t control.
Open models like DeepSeek V4 and GLM-5.2 are closing the gap faster than everyone expected. They’re not perfect but they don’t come with a citizenship requirement either. At least for now. It’s probably a good time to start paying attention.
Fun times.
Edit: As of July 1st 2026 the export controls for Fable 5 and Mythos 5 have been lifted, but the damage is already done