Can the U.S. government legally gatekeep global access to AI models?

| Source: Fast Company AI

Tags: export controls, AI policy, Claude, Anthropic, trade law, regulation, U.S. government

For the first time, the U.S. government used export controls to disable Claude Fable 5 and Mythos 5 for foreign users — blocking international access to specific AI services and raising urgent legal questions about Washington's authority to restrict global AI access under existing trade law.

Details

The shutdown of Claude Fable 5 and Mythos 5 for foreign users represents a historic first: the U.S. government applying export control law to cut off international access to AI models-as-a-service. While export controls have long covered hardware and software with military applications, applying them to AI inference APIs is uncharted territory under existing statutes like the Export Administration Regulations (EAR). The core legal question is whether the government has clear statutory authority to treat access to an AI API as an export of controlled technology. Current export control frameworks were built around physical items and traditional software, not cloud-hosted models accessible over the internet. Legal analysts and AI companies are now scrambling to understand what capability threshold triggers export restrictions — and whether such controls can be enforced against foreign users of cloud services. The practical implications are broad. If this precedent holds, any sufficiently capable AI model hosted by a U.S. firm could face export licensing requirements for foreign users — reshaping how American AI labs deploy globally and potentially accelerating foreign investment in non-U.S. AI infrastructure. The Fast Company article raises these questions but is light on specific statutory citations or government statements, making the full legal picture unclear.