The United States announced an interim rule that will require firms to obtain licenses before exporting artificial intelligence (AI) chips if the chips could be used to provide compute services to China or other restricted countries. The rule, issued by the Commerce Department's Bureau of Industry and Security (BIS), is intended to prevent foreign entities from accessing advanced compute power remotely, a capability that could accelerate the development of sophisticated AI models.
The move marks the latest escalation in a series of U.S. measures aimed at curbing the flow of high-performance semiconductor technology to strategic competitors. By targeting not only the physical shipment of chips but also the provision of cloud-based compute, the rule expands the scope of export controls to cover a broader range of commercial practices.
What the New Interim Rule Requires
Under the interim rule, any U.S. firm that exports AI chips classified under the Export Administration Regulations (EAR) and that knows, or has reason to know, that the chips could be used to deliver remote compute services to a restricted destination must first secure an export license from BIS. The rule applies to chips that are capable of supporting large-scale AI training workloads, typically those with high memory bandwidth, large numbers of tensor cores, or specialized accelerators designed for deep-learning inference.
Companies must submit detailed end-use statements, describing the intended customers, the nature of the compute services, and the safeguards in place to prevent unauthorized re-export. BIS will evaluate each application based on national security criteria, including the potential for the technology to be used in military or surveillance applications.
Why the United States Is Targeting Remote Compute
Remote compute services allow a user in one country to run AI workloads on hardware located in another country, often via cloud platforms. This model has become popular because it lets firms and research institutions access cutting-edge chips without the capital expense of purchasing and maintaining the hardware themselves.
U.S. officials argue that allowing Chinese companies to tap into American-made AI chips through remote services effectively circumvents existing export bans. "If a chip can be shipped to a foreign party, the risk is clear," said a senior BIS official. "If the same chip is used to provide compute over the internet, the risk is equally serious, and our current rules did not address that loophole."
Historical Context of U.S. AI Export Controls
Export controls on advanced semiconductors date back to the Cold War, when the United States restricted the flow of certain technologies to the Soviet Union. In recent years, the focus has shifted to China, which has invested heavily in AI research and aims to become a global leader by 2030.
In October 2022, the U.S. added several Chinese AI firms to the Entity List, requiring licenses for any transaction involving U.S. technology. In 2023, the Department of Commerce expanded the definition of "high-performance computing" under the EAR, capturing a wider array of GPUs and AI accelerators. The current interim rule builds on those precedents by addressing the service-based model of compute delivery.
How the Rule Affects Companies and Researchers
Technology firms that sell AI chips to overseas distributors will need to revise their compliance programs. This includes updating sales contracts, implementing new screening procedures for end-users, and possibly redesigning cloud-service offerings to ensure they do not fall under the new licensing requirement.
Academic institutions that rely on cloud-based AI training for collaborative projects with Chinese partners may also be impacted. Researchers will need to verify that any remote compute resources they use are sourced from providers who have obtained the necessary licenses.
Enforcement and Penalties
Violations of the interim rule could result in civil penalties of up to $1 million per violation, as well as criminal fines and possible export bans. BIS has indicated that it will conduct audits of firms that export AI chips and will cooperate with other agencies, including the Department of Justice and the Federal Bureau of Investigation, to investigate suspected breaches.
What Comes Next
The interim rule will be in effect for a 180-day comment period, after which the Commerce Department may issue a final rule. Stakeholders have been invited to submit written comments, data, or case studies that illustrate the rule's impact on legitimate trade and research.
In the meantime, companies are advised to conduct an internal review of all AI chip export activities and to engage legal counsel experienced in export compliance. Organizations that already provide remote compute services to China should assess whether a license is required and, if so, begin the application process promptly.
Watch for further guidance from BIS, which is expected to publish detailed licensing forms and FAQs within the next few weeks. The rule's implementation will also be monitored by congressional committees overseeing technology and national security, potentially leading to additional legislative action.
Overall, the policy reflects a broader U.S. strategy to maintain a technological edge while limiting the ability of rival nations to harness American innovations for military or surveillance purposes.