The U.S. House of Representatives passed the Remote Access Security Act with an overwhelming majority on January 12th, expanding export controls from physical AI chips to cloud computing services. The aim is to prevent Chinese companies from circumventing U.S. chip restrictions through remote means. Experts suggest that this move signifies an extension of the U.S.’s technological control over China from hardware restrictions to “full-chain management of computing power,” which may slow down China’s development in artificial intelligence.
On January 12th, the U.S. House of Representatives passed the Remote Access Security Act with a vote of 369 to 22. The bill was introduced by Republican Congressman Mike Lawler from New York and garnered widespread support from both Democratic and Republican parties.
This legislation expands the control scope of the 2018 Export Control Reform Act, stipulating that if remote access to controlled items poses a serious threat to U.S. national security or foreign policy, it will be included in the control range.
According to the text, the controlled remote access includes: (1) training artificial intelligence models that can significantly lower the threshold for designing or using weapons of mass destruction or automatically discovering and exploiting vulnerabilities for network attacks; (2) accessing quantum computers that could aid in network attacks or pose other national security risks; (3) obtaining hacking tools.
Congressman Lawler stated in a press release that the current export control laws have serious loopholes: companies allied with the Chinese Communist Party can access restricted U.S. technology through cloud services. Currently, China is bypassing export control laws through cloud services. This bill clearly defines “remote access,” authorizing the Department of Commerce to regulate controlled technologies accessed through the cloud to effectively close this security gap.
Chairman of the House Committee on China Competition Strategy, Republican Congressman John Moolenaar from Michigan, stated in a declaration that this bill “brings export control laws into the digital age” and specifies that relevant regulations apply equally to cloud computing services and physical chips.
Xie Peixue, Deputy Researcher at the Network Security and Decision-Making Simulation Research Institute of the Taiwan National Defense Research Institute, indicated to Epoch Times that the bill clearly conveys the U.S.’s policy direction of expanding technological control from physical products to virtual computing power. “Even though it may not completely block it, it is enough to significantly increase the cost for China (CCP) to access top AI computing power.”
Wang Xiaowen, Assistant Researcher at the CCP Military and Operational Concept Research Institute of the Taiwan National Defense Security Research Institute, holds a similar view. She stated to Epoch Times that the key to artificial intelligence competition lies not only in the chips themselves but in the sustainable access to overall computing power and advanced level.
Concerns about cloud computing potentially undermining the effectiveness of export controls are not new. The Wall Street Journal reported in July 2023 that the Biden administration was considering requiring U.S. cloud service providers to obtain government permission before providing services using advanced AI chips to Chinese customers.
Researcher Emily Weinstein from the Georgetown University Center for Security and Emerging Technology pointed out at the time that even if Chinese companies cannot directly purchase Nvidia A100 chips, they can legally use related computing power through cloud services, effectively weakening the constraints of export controls.
In September 2024, the previous U.S. House of Representatives passed a similar version of the Remote Access Security Act, but it did not complete the legislative process in the Senate. This recent passage indicates ongoing congressional attention to plugging related loopholes.
Cloud computing is also considered a key area of competition between the U.S. and China. The Financial Times of England reported in 2021, citing Atlantic Council researcher Justin Sherman, that cloud computing will become a crucial battlefield for U.S.-China competition in the coming years. According to Canalys data, China’s cloud computing spending in 2020 was about $19 billion, accounting for approximately 13% of the global market, second only to the United States.
The House version of the Remote Access Security Act has been submitted to the Senate for review. The Senate version of the bill was introduced in December 2025 by Republican Senator Dave McCormick and Democratic Senator Ron Wyden. In accordance with U.S. legislative procedures, the bill must pass both chambers in a consistent form before being sent to the President for signature and enactment.
Wang Xiaowen indicated that if the bill smoothly passes and is implemented, it may slow down China’s pace in developing large-scale models and further widen the development gap between the U.S. and China in the field of artificial intelligence.
She believes that with limited production capabilities for high-end AI chips in China, cloud computing has become an important alternative path. Although complete blockage is difficult, the U.S.’s dominant position in computing power provides leverage strategically, and these measures could be seen as important bargaining chips in future negotiations with China.
Xie Peixue noted that the bill’s direct impact is in restricting Chinese companies from using overseas cloud space to “circumvent” physical chip regulations. “Without cloud restrictions, Chinese companies may still indirectly use more advanced computing power through U.S. cloud platforms,” he said. U.S. cloud services not only provide computing power but also integrate key software ecosystems like CUDA. If Chinese companies lose access to these platforms, they will have to build their own complete hardware and software systems, leading to significantly higher costs and time for small and medium-sized AI enterprises.
“In a situation where high-end chip and cloud computing capabilities are both constrained, Chinese companies may be forced to rely on existing inventory or domestic chips for model training, putting pressure on training efficiency and costs,” he added.
However, Xie Peixue also cautioned that the execution of the Remote Access Security Act faces practical challenges, including companies establishing subsidiaries in third countries, using VPNs to evade identity detection, as well as cross-border data and privacy legal issues. He described this as potentially evolving into a long-term “enforcement and evasion” game.
While the U.S. is pushing for legislation on cloud computing power, it allowed the export of Nvidia’s H200 chip to China. Xie Peixue believes this reflects the differences in the division of labor between the U.S. executive and legislative branches in their technological policies towards China.
He analyzed that under the system of checks and balances, the executive branch often seeks flexibility between national security and corporate interests, while the Congress generally adopts a tougher stance on China-related issues and tends to “fill institutional gaps” through legislative means. This legislation can be seen as a “hedge and reinforcement” of the policies of the executive branch.
He also mentioned that the H200 is no longer cutting-edge technology. With Nvidia’s new B200 architecture chips entering the market, the significant gap between the U.S. and China in high-end AI computing power remains evident.