U.S. Administration Weighs Restrictions on Access to Chinese AI Models

The Trump administration has discussed measures that could limit U.S. access to advanced Chinese artificial intelligence models. Such steps would require licenses for certain technologies and could affect companies using open-source systems developed in China.

U.S. firms have adopted Chinese open-source models due to lower costs and performance levels comparable to domestic alternatives, including the recent Kimi system. Parts of the administration previously examined ways to restrict foreign open-source models, according to individuals familiar with internal discussions.

The Commerce Department considered adding several Chinese AI laboratories to the Entity List in 2024, which would have required U.S. entities to obtain licenses before accessing their technology, a source close to the administration said. The National Security Agency and the White House Office of the National Cyber Director also reviewed options for issuing guidance on potential risks associated with Chinese AI laboratories, the same source stated. The White House examined an executive order that would have permitted U.S. companies to host Chinese models only if they could certify security standards and accept liability for any breaches, according to the source.

In summer 2024, Commerce circulated draft rules that would have used existing authorities on domestic supply chains to address Chinese open-source models, another individual familiar with the process said. Officials focused on avoiding regulatory constraints on innovation opposed those proposals at the time. Personnel changes, including the departure of former White House adviser Sriram Krishnan, have occurred since then.

The United States added multiple Chinese entities to the Entity List during 2025, citing national security concerns related to artificial intelligence and computing capabilities. These designations provide documented precedent for restricting technology transfers in this sector.

Current discussions have examined alternatives to outright prohibitions. One approach involves procurement requirements, Entity List designations, and public statements directed at companies using Chinese models, according to a source familiar with government deliberations. Another described effort centers on identifying potential security vulnerabilities in Chinese systems and encouraging development of U.S. open-source alternatives, the source added.

David Sacks, an external adviser to the White House on artificial intelligence, posted on X that leading closed laboratories hold a dominant position in model revenue and seek government action against open-source competition. Sacks has previously cautioned against regulatory outcomes that would favor the largest U.S. laboratories. Individuals close to the administration indicated that laboratories or affiliated parties have raised proposals regarding open-source models at intervals of three to five months.

Chinese models continue to offer cost and practicality advantages relative to available U.S. alternatives in some applications. Neither the White House nor the Commerce Department provided comment on the reported discussions.

Chinese leader Xi Jinping has emphasized artificial intelligence development. U.S. industry participants face decisions on technology adoption amid ongoing policy considerations. Anthropic and OpenAI have advocated for safety frameworks that include licensing elements in certain cases.