The nightmare for AI companies is not just regulation. It is being told to shut off a product globally because the government cannot comfortably fit it inside old export-control rules. Legion LegalTech Corp. says that is exactly what happened when a June 12 directive forced Anthropic to disable access to Fable 5 and Mythos 5 for foreign nationals worldwide.
- Legion LegalTech Corp. sues to vacate a federal directive forcing Anthropic to disable model access for foreign nationals.
- The June 12 order targets Fable 5 and Mythos 5, impacting millions of commercial users across global software development pipelines.
- The lawsuit challenges whether the Bureau of Industry and Security can regulate hosted AI services under traditional physical export-control statutes.
The Order
According to the complaint, the Bureau of Industry and Security told Anthropic to immediately disable access to the models for any foreign national anywhere in the world, under threat of criminal and civil penalties. Anthropic complied the same day, saying it could not reliably apply real-time nationality filtering across its products without shutting the models off altogether.
Legion says the result was immediate harm. It uses Fable 5 in AI-powered drafting and case-management tools for lawyers, and says the shutdown disrupted a core part of its development workflow.
The Claims
Legion argues the government exceeded its export-control authority. The complaint says current rules do not clearly classify access to a hosted AI model, or the text it produces, as a controlled item.
The lawsuit also invokes the Berman Amendment, which limits the government’s power under emergency economic law over informational materials. Legion says the outputs it receives from the model, including legal analysis, summaries and drafted text, fall within that protection.
Have a development worth tracking?
Share product launches, funding announcements, partnerships, research findings and market developments with The Grey Terminal's readership.
→ Submit a Press ReleaseIt further argues the directive was arbitrary because the alleged risk was narrow, yet the remedy was a global suspension that swept in allies, employees and commercial users far beyond the supposed threat.
Broader Context
The case lands at a sensitive moment for AI policy. The administration has been pushing a tougher posture on frontier model controls, even as its June 2 executive order said the government should not create a mandatory licensing or preclearance regime for new AI models.
That tension gives the lawsuit broader significance than a single customer dispute. At issue is whether the government can treat access to a hosted AI model like a regulated export in the same way it would chips, software or controlled technical data.
The answer may shape how frontier AI services are sold, deployed and governed. If the court accepts the government’s theory, hosted models could become subject to a much tighter compliance regime than the industry has assumed.
Next Steps
Legion is seeking to vacate the directive and block enforcement through an injunction. The government is likely to lean on national security arguments and the deference courts often give export-control decisions.
Even so, the case may become an early test of how far regulators can stretch traditional trade rules to cover live AI services. It also raises a practical question the industry has not fully answered: whether cloud-based models can be restricted by nationality at all without breaking the product.
Grey Terminal Note
This case is bigger than one company’s access to one model. It is a test of whether the law can treat hosted AI like an export, or whether the legal framework still assumes software moves like a shipment rather than a service that answers in real time.
Activate Terminal Layer
Structural analysis of the systems, pressures, and stakeholders behind this story.





