The White House this week formalized a new posture on artificial intelligence, seeking to wall off Chinese competitors from American intellectual property while preserving the open-weight ecosystem that domestic startups and researchers depend upon. The move draws a clear line between legitimate refinement of models and industrial-scale theft, laying the foundation for sanctions and Entity List designations against Beijing's AI labs.
Two Tiers of Distillation
Michael Kratsios, chief technology adviser to President Trump, articulated the administration's two-tier standard. "Legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem," Kratsios stated. He immediately contrasted that with conduct the White House views as hostile: large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research. The distinction allows policymakers to armor-plate domestic IP without throttling the open-source models that power American firms and university labs.
Chinese Labs in the Crosshairs
The immediate target is China's Moonshot AI. Administration officials accused the firm of using distillation to effectively clone Anthropic's Fable model through fraudulent accounts and evasive access techniques. Treasury Secretary Scott Bessent abandoned ambiguity on consequences. "When PRC firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table," Bessent said. Adding a Chinese AI lab to the Commerce Department's Entity List would sever U.S. access without a license, a hammer American workers have seen wielded successfully against other sectors.
Economic Nationalism Meets Tech Policy
The administration's framing threads economic nationalism directly into AI governance. American taxpayers and private investors fund the enormous compute costs behind frontier models, and the White House is signaling that foreign state-linked enterprises will not be permitted to siphon that value without consequence. The U.K. Artificial Intelligence Security Institute and the U.S. Center for AI Standards and Innovation released a joint evaluation Thursday finding Moonshot's Kimi K3 performs significantly below other frontier models in cyber capabilities—an assessment that reinforces the administration's narrative that Chinese labs are subsidizing mediocrity with stolen methods while undercutting American-market alternatives on price.
Open Development Preserved
Industry figures confirmed the White House has not sought to ban Chinese open models outright. OpenAI co-founder Greg Brockman told reporters he had not been involved in any such conversations, characterizing distillation's challenges as primarily technical. The White House appears to be betting it can preserve open-weight competition for American developers while using tariff and sanctions levers to prevent adversaries from leapfrogging on the backs of U.S. research. For domestic workers and firms that rely on efficient, specialized models, the path forward remains intact—provided the underlying technology wasn't stolen through what the administration now unambiguously labels industrial espionage.