Anthropic's Sudden Shutdown Exposes the Chaos in AI Export Policy

Last week, the Trump administration hit Anthropic with an abrupt export control order that forced the company to cut off access to its Fable 5 and Mythos 5 models for all foreign nationals - including users inside the U.S. and even some of its own employees. The move, which the government justified only as a "national security" directive, has left the AI industry scrambling for answers.

Unlike traditional export controls on hardware or software, this order targets a cloud-hosted AI service. Users don't download model weights or source code; they simply query a chatbot. Experts say this stretches the legal framework to a breaking point. "To my knowledge, this is the first time US export controls have been used to control access to an AI model in this way," said Hanna Dohmen of Georgetown's Center for Security and Emerging Technology.

Andrew Reddie, a public policy professor at UC Berkeley, noted that while the government has wide latitude under existing rules, "the equivocation by successive administrations regarding what the responsibilities of model developers are" has created dangerous uncertainty. If Anthropic was targeted because of its models' capabilities, what happens when OpenAI or Google launch their next-generation systems? If it was about security flaws, the government needs to define acceptable safeguards. And if it was simply political, the industry has no roadmap for compliance.

Either way, this ad hoc approach is unsustainable. "If creating models that are impossible to jailbreak becomes the de facto standard for the United States, then it will have no AI models," Reddie warned. The episode is already fueling arguments abroad that relying on American AI firms is risky. For the U.S. to maintain its lead, Washington must replace opaque, last-minute orders with a transparent, predictable governance framework.

Source: The Verge