The current U.S. export controls on semiconductor chips, particularly those used in advanced AI, are broad and apply to any buyer in China, even for benign commercial uses. These controls not only affect U.S. AI chips but also chips produced outside the U.S. using U.S.-origin technology. This approach risks damaging the market dominance of U.S. firms and may not effectively slow the proliferation of high-risk AI systems, as prohibited actors can still access chips through smuggling or cloud computing resources.

Cloud computing can provide a more flexible and effective means of regulating access to AI chips. By allowing specialized AI chips to be accessed through secure cloud infrastructure, it becomes easier to monitor usage and enforce compliance with regulations. This method enables better detection of misuse, improves cybersecurity, and allows for targeted controls that can adapt to changing circumstances, ultimately preserving the competitiveness of U.S. chip firms while mitigating risks associated with high-risk AI models.

Governing AI chips through cloud infrastructure is crucial for U.S. national security as it allows for better oversight of who can access advanced AI capabilities. By maintaining visibility over AI compute access, the U.S. can prevent malign actors from misusing powerful AI systems while still enabling commercial uses that do not pose risks. This approach helps to ensure that U.S. firms remain competitive in the global market and reduces the likelihood of adversaries developing their own advanced AI capabilities through unrestricted access to AI chips.