The Trump administration is caught in a self-inflicted policy collision. It created a major initiative to flood global markets with American AI systems, but its own export restrictions are now sabotaging that very strategy.
Last month, the White House slapped export controls on Anthropic's Claude 5 model, blocking the startup from selling it overseas due to safety disagreements. Anthropic responded by yanking the model from international access entirely. The two sides remain locked in talks this week with no resolution in sight.
The fallout goes beyond one company. The move exposes a fundamental contradiction at the heart of Trump's AI agenda: a program designed to dominate global AI markets cannot function if the government arbitrarily cuts off access to America's leading models.
Dean Ball, who advised the Trump administration on AI policy, put the problem bluntly. "The government's willingness to arbitrarily and abruptly remove America's best models from all foreign use shows that the strategy behind the AI Export Program is no longer relevant to decision makers in the U.S. government," he told Axios.
The American AI Export Program itself is barely six months old. Trump created it through executive order in July 2025 as a centerpiece of his administration's AI dominance strategy. Participating companies get expedited export licenses, priority access to federal credit, direct government support with foreign governments, and coordinated interagency help. It was supposed to be a competitive advantage for U.S. tech firms.
But industry insiders say the Anthropic move has poisoned the well. Tech executives worry that joining the program means nothing if Washington can reverse course at any moment.
The downstream damage extends beyond lost sales. One industry source warned that using export controls as an enforcement tool for tech policy sets dangerous precedents. "Fueling perceptions that the US government could disable overseas access to an AI model or system only makes it more difficult to promote American AI exports," the source said. "Global customers will have a harder time committing to purchasing US-made AI."
Allies are particularly concerned. The restrictions complicate relationships at a moment when the White House is pushing hard to export American technology to partners abroad. One tech source flagged another risk: AI components are interdependent. Block one layer and you can disrupt the entire stack.
Paul Lekas, vice president of public policy at the Software and Information Industry Association, raised an even sharper concern. "It definitely has a flavor of picking winners and losers," he said, warning that companies want consistency, not favoritism that varies company to company.
The White House says the move balances innovation with national security, and spokesman Kush Desai defended the action on those grounds. The Commerce Department declined to comment.
Some analysts suggest this could be a one-off dispute that gets resolved without repeating. Monument Advocacy's Joseph Hoefer noted that the upside of joining the export program is still attractive, but companies now have to plan contingencies for sudden access cuts. Applications for the program are due June 30, and how the administration resolves its fight with Anthropic will likely determine whether other AI firms feel confident enough to participate.
Author James Rodriguez: "You cannot build a global export strategy on quicksand, and using export controls as a political cudgel undermines the entire enterprise."
Comments