• Download a copy of the lawsuit here. 

The US government cannot use export-control laws to restrict access to Anthropic’s AI models because no existing export control covers hosted AI models or their outputs, according to a lawsuit filed by Legion LegalTech against the US government in a federal court in Washington, DC.

Filed on June 23, the lawsuit argues that the government’s June 12 directive ordering Anthropic to disable its Fable 5 and Mythos 5 models for all foreign nationals worldwide exceeded its statutory authority under US export-control laws and emergency powers legislation. The legal technology company, which relied on Fable 5 for product development and employed Canadian developers, says the restrictions disrupted its operations and prevented its engineers from accessing tools central to its business.

The lawsuit seeks a declaration that the directive is unlawful, an order setting it aside, and an injunction preventing its enforcement. Legion argues that the government acted beyond its legal authority, misapplied export-control provisions, violated limits contained in the International Emergency Economic Powers Act (IEEPA), and issued the directive without a lawful basis.

No export control covers hosted AI models: The lawsuit argues that the government’s export-control authority does not extend to access to hosted AI models.

According to the filing:

• The “only export-control classification that ever directly covered advanced AI model weights — ECCN 4E091 — was rescinded in May 2025 with no replacement”.
• “No currently operative provision of the Commerce Control List classifies access to a hosted AI model, or its inferential text output, as a controlled item.”

The lawsuit therefore alleges that “Commerce cannot enforce a control that does not exist.”

The complaint further contends that users of Fable 5 never received model weights, source code, or technical data and therefore did not receive anything that would qualify as a controlled export.

Government stretched export-control powers beyond intended purpose: Legion also challenges the legal mechanisms the government relied on to suspend access to Anthropic’s models.

According to the complaint:

• The directive relied on Section 4817(b)(1) of the Export Control Reform Act and a BIS “Is Informed” letter issued to Anthropic.
• Legion argues that Section 4817(b)(1) can only be used after the government follows a formal process, including rulemaking and coordination with international partners, which “never occurred.”
• The company also argues that “Is Informed” letters are normally used in specific cases where authorities identify concerns about a particular customer, end user or military-related activity.
• However, the June 12 directive applied to “any foreign national” worldwide, regardless of nationality, location or intended use.
• The complaint describes the measure as a blanket restriction and argues that the government used a “targeted, case-specific tool for identified military-intelligence end-use risks” as “a mechanism for imposing blanket worldwide bans”.

As a result, Legion claims the government acted “in patent excess of statutory authority” by using export-control provisions in a manner not authorised by law.

IEEPA cannot be used to restrict AI outputs: The complaint further argues that the directive is unlawful if it relies on the IEEPA.

According to the filing:

• The Berman Amendment excludes from IEEPA authority any power to regulate “the exportation of information or informational materials in any format or medium of transmission”.
• The statute, the lawsuit says, does not merely limit such authority but “withholds it entirely”.

Legion argues that AI-generated outputs, including “drafted text, written legal analysis, summaries, and similar composed material,” qualify as informational materials protected by the exemption.

The complaint therefore alleges that the directive’s purpose and effect are to stop the flow of informational materials to users.

No qualifying national emergency

According to the complaint:

• “No national emergency has been declared with respect to the purported threat” posed by the models.
• The alleged threat did not have its “source in whole or substantial part outside the United States” as required under the statute.

The filing notes that the models were developed by a US company, hosted in the US, and offered through a US commercial platform.

Directive conflicts with White House policy: The lawsuit additionally points to a June 2 executive order signed by President Donald Trump, 10 days before the Anthropic directive.

According to the complaint:

• The order stated that nothing should be interpreted as authorising “a mandatory governmental licensing, preclearance, or permitting requirement for the development, publication, release, or distribution of new AI models.”

Legion argues that the June 12 directive imposed “precisely such a requirement.”

The filing states: “Where the Executive publicly disclaims a power and then exercises it, the inference that no statute confers the power is compelling.”

Arbitrary and capricious action: Finally, the complaint alleges that the directive was arbitrary in nature:

• The government cited “a narrow jailbreak allowing a model to review software code” but imposed “a categorical worldwide suspension,” which it argues is “materially underinclusive.”
• The restrictions applied to all foreign nationals regardless of nationality or location.
• Similar capabilities remained available through competing products, including OpenAI’s GPT-5.5.
• The government failed to consider “obvious less-restrictive alternatives,” including limiting restrictions to adversary nations.

Also read:

• SEBI forms task force, orders immediate cybersecurity overhaul amid Claude Mythos concerns
• Anthropic calls for the ability to slow or pause frontier AI development
• Anthropic opens Mythos-class AI to the public, but keeps high-risk capabilities behind guardrails