1 hour ago
1
An AI startup built its business on Anthropic’s most powerful models. Then the US government turned them off.
Legion Intelligence, a company focused on national security applications, has filed a lawsuit against the United States government after a Commerce Department directive effectively killed its access to Anthropic’s advanced AI systems. The case represents the first major legal challenge to export controls being applied to AI models, and it’s forcing a question nobody in Washington seems eager to answer: what happens when national security tools get classified as national security threats?
What happened
On June 12, 2026, US Commerce Secretary Howard Lutnick issued an export control directive targeting Anthropic specifically. The order prohibited access to two of the company’s most capable models, Fable 5 and Mythos 5, for all foreign nationals. Not just foreign nationals abroad. Foreign nationals inside the United States, too.
Anthropic complied almost immediately. By June 13, the company had disabled access to both models across its entire customer base to meet the directive’s requirements.
The legal basis for the order sits in the Export Control Reform Act of 2018, a law originally designed to regulate the transfer of sensitive technologies to foreign adversaries. This marks the first time that law has been invoked to restrict access to AI models specifically. The directive cited national security risks tied to potential misuse by foreign military intelligence services, with China and Russia named as primary concerns.
For Legion Intelligence, the timing was brutal. The company had built its product stack around Anthropic’s technology, and the blanket restriction didn’t distinguish between a Chinese military researcher and a Canadian software engineer with a green card working at a US defense contractor.
Legion pushes back
Legion CEO Ben Van Roo didn’t wait long to make his position clear. On June 23, he publicly criticized the directive’s approach, calling the “no foreign national” rule “the most impossible thing to enforce.”
“The most impossible thing to enforce.”
Legion’s lawsuit argues that the blanket restriction goes beyond what national security requires and imposes an unconstitutional burden on companies that were, ironically, building tools designed to help the US national security apparatus.
The Anthropic angle
This situation gets more complicated when you zoom out on Anthropic’s relationship with the federal government. The company has been embroiled in its own disputes with the Pentagon since at least March 2026, when lawsuits emerged over Anthropic’s designation as a supply chain risk.
Anthropic’s decision to comply with the directive within 24 hours left customers like Legion stranded overnight, with no transition period and no alternative access.
What this means for investors
The Export Control Reform Act gives the Commerce Department broad authority to restrict technologies deemed sensitive to national security. This marks the first time it has been pointed at AI models specifically.
Companies building on top of Anthropic, OpenAI, Google, or any other frontier AI provider now face a dependency risk tied to geopolitical exposure. If a startup is selling AI-powered analytics and the Commerce Department decides its model provider’s latest release is too sensitive for foreign nationals, its revenue disappears instantly.
Chinese AI companies like DeepSeek and Alibaba’s Qwen have been aggressively expanding their model capabilities. US export controls on American AI models don’t just restrict foreign access. They create market opportunities for non-US alternatives that face no such restrictions. Every customer locked out of Anthropic’s Fable 5 is a potential customer for a Chinese model that works just fine without a citizenship check.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) · 