A US startup is taking the American government to court after a forced access cut to Anthropic’s Fable 5 left its product development stranded. Legion, which builds AI tools for legal workers, says the order does not just slow it down – it threatens the company’s survival, and it wants a federal judge in Washington to say the restriction overreached.
The fight comes less than two weeks after US authorities told Anthropic to limit access to its advanced Mythos 5 and Fable 5 models for foreign citizens. That move may have looked like a narrow compliance step on paper, but in practice it has created a new fault line for AI companies that rely on distributed teams. The tighter the access rules, the easier it is for a regulator to hit a startup’s workflow with one order and a lot of collateral damage.
Why Legion says the access cut is existential
Legion’s case is unusually awkward for Washington because the company is formally based in the US but leans on software developers in Canada. Under the new restrictions, that setup is enough to cut off access to Fable 5, which Legion says was in active use during development. The startup argues that every day without the model weakens its chances of making it through the crunch.
There is also a broader industry tension here: AI firms have spent the last few years pitching borderless teams and cloud-delivered tools as the future, while governments are increasingly treating advanced models like sensitive infrastructure. That clash is hardly unique to Anthropic, and competitors such as OpenAI are watching the precedent closely because a similar order could land on them with very little warning.
Howard Lutnick is named in the complaint
The defendant named in the lawsuit is US Commerce Secretary Howard Lutnick, who issued the directive blocking foreign citizens from Mythos 5 and Fable 5. Legion’s co-founder and chief executive, Arthur Rothrock, says the suit is about more than one model or one vendor. In his view, letting the government cut off access this way creates a dangerous legal template for future action against any AI company that depends on internationally spread talent.
Anthropic, for its part, has publicly kept its response careful and brief, thanking the government for trying to sort out the problem quickly. That is the kind of corporate politeness that usually shows up when no one wants to say too much while lawyers are sharpening pencils behind the scenes.
What the Anthropic access cut could mean for other AI startups
For now, Legion is the test case. If it wins even part of its argument, startups with cross-border engineering teams may get a little more room to push back against model access rules. If it loses, the message is blunt: build globally if you want, but don’t assume the model powering your product will stay available when politics shifts.