China Select Committee Backs Host of Export Control Bills to Keep US AI Lead

In a new report on China’s efforts to improve its AI capabilities, the committee asserted that the legislation it endorsed would counter the mix of legal and illegal means China uses to acquire American technology.

“China is willing to buy what they can, and steal what they cannot, to advance their AI ambitions,” committee Chairman John Moolenaar, R-Mich., said. “The task for Congress is to pass legislation that will stop China’s multiprong effort to legally and illegally acquire America’s tech stack to use it against us.”

The committee said the AI Overwatch Act, the Remote Access Security Act and the Scale Act would close loopholes that allow Chinese firms to legally access advanced U.S. chips and cloud computing. The AI Overwatch Act would increase congressional oversight of sales of advanced AI chips to China, while the Remote Access Security Act would authorize the Bureau of Industry and Security to regulate controlled items through the cloud (see 2601210037 and 2601130006). The Scale Act, which had not been previously disclosed, would set export limits based on China’s production capacity.

The committee also called for passage of the Chip Security Act, which would require anti-diversion measures for export-controlled advanced chips; the Match Act, which seeks to tighten and harmonize U.S. and allied controls on sales of chipmaking tools to China; and the Stop Shells Act, which would place subsidiaries on the BIS Entity List or Military End-User List if they are owned 50% or more by companies on those lists (see 2603260058, 2604140004 and 2510010011).

Of the six named bills, the Remote Access Security Act is farthest along in Congress, having passed the House in January (see 2601130006). It awaits Senate consideration.

The China Select Committee's other recommendations for lawmakers included requiring a BIS license to provide cloud access to controlled AI chips; directing BIS to eliminate certain exceptions in its Foundry Due Diligence Rule; calling for the investigation and prosecution of unauthorized copying or “distillation” of U.S. AI models’ capabilities; encouraging BIS to consider adding Chinese AI firms DeepSeek, Moonshot AI and MiniMax to its Entity List; and increasing civil and criminal penalties for export control violations.

During a hearing on China’s AI efforts, Dmitri Alperovitch, chairman of the Silverado Policy Accelerator, said U.S. export controls “have prevented China from closing the compute gap” but should be enhanced. He would avoid allowing the sale of advanced AI chips to China, which he compared to “selling rockets to the Soviets during the moon race.”

Alperovitch would also ban the sale of advanced chipmaking tools to China and strengthen protections against AI model distillation attacks. He would create know-your-customer rules for cloud companies to ensure Chinese AI models don't get trained by chips outside China.

Kyle Chan, fellow at the Brooking Institution’s John L. Thornton China Center, told the committee that export controls “are not a panacea, but they have bought us some time. The question is how we use this time effectively.” In written testimony, Chan said he would complement export controls with "sustained investment in infrastructure, research, and open ecosystems."

Yusuf Mahmood, director of AI and emerging technology at the America First Policy Institute, offered several recommendations to secure U.S. AI, including providing $50 million to $100 million annually to the Commerce Department’s recently renamed Center for AI Standards and Innovation, or CAISI. He would also authorize the State Department’s new Bureau of Emerging Threats. “These offices are critical for the situational awareness needed to monitor and counter” China, Mahmood said.