No Confirmation of Proposal to Require U.S. Companies to Disclose Use of Chinese AI

Key Facts
The report that President Donald Trump’s administration is considering requiring U.S. companies to disclose their use of Chinese AI models could not be verified, based on an assessment of the available evidence as of November 18, 2026. The circulating report described a proposed policy focused on disclosure of the use of those models, but no reliable open-source material available confirms the proposal’s existence or details. It is therefore not possible to conclude that Washington intends to impose a new requirement on companies or determine its scope. There is also no evidence that the proposal, if it exists, would ban the use of Chinese models or impose direct restrictions on them. The issue comes amid a broader U.S. debate over China’s access to advanced chips and AI-related infrastructure.
Against the backdrop of this debate, The Associated Press said proposals under consideration in Congress include giving lawmakers a role in reviewing licenses to sell advanced chips to China. The proposals also include tracking chips to prevent them from being diverted to unauthorized destinations, as well as pressing U.S. allies to align export restrictions. The agency reported that a new rule requires data centers outside the United States that serve companies in countries including China, Russia, Iran and Cuba to identify end users. The rule also adds licensing requirements for some chip sales, according to the agency. These measures highlight a regulatory track involving chips and data centers, separate from requiring companies to disclose which models they use.
Bloomberg reported in August that U.S. companies, including Airbnb, DoorDash and Coinbase, use Chinese AI models to a limited extent. It quoted an Airbnb spokesperson as saying the company runs a limited number of open-source Chinese models after they are adopted by U.S. service providers that separate data and operations and keep them protected. The example highlights that using the models does not necessarily mean transferring data or operations to a Chinese provider, according to the company’s description of its arrangements. Bloomberg said companies differ in their assessments of the security risks associated with these models. The debate thus spans how companies use these tools and concerns about data security and the code they produce.
In terms of usage, Bloomberg said Chinese models’ share of U.S. token traffic surpassed that of U.S. models for the first time earlier in the year, reaching 63% in the first week of July 2026. Token traffic refers to the amount of data processed by AI models, according to the report. The figure indicates the growing use of Chinese models, but on its own does not establish how many companies rely on them or the nature of the tasks they perform. Separately, Bloomberg quoted researchers at Booz Allen as saying some Chinese models produced less secure code in certain cases, potentially threatening the confidentiality of information. These concerns remain part of the ongoing security debate, but none of them establishes that the disclosure proposal attributed to the administration exists.
The disclosure debate is taking place alongside parallel congressional moves concerning chip exports. On November 12, 2026, Representatives Ron Wyden and Richard Neal asked the Commerce Department to preserve records related to semiconductor investigations, Nvidia export license approvals, and the department’s communications with the company and its officials, according to a letter from the Senate Finance Committee. The letter concerns investigations, licenses and official communications, and in itself does not provide evidence of a policy requiring companies to disclose their use of Chinese AI models. On the U.S.-China front, Reuters said tighter restrictions on China’s access to advanced chips are likely to overshadow talks between Presidents Donald Trump and Xi Jinping at this week’s Shenzhen summit. It added that officials from both sides planned additional discussions on AI on the sidelines of the summit, addressing safeguards against misuse of the technology.