
Two powerful House chairmen called on Friday for Chinese-developed artificial intelligence models to be removed from American federal systems and high-risk critical infrastructure, concluding a seven-month investigation that found major US companies — including Airbnb, DoorDash and the maker of the Cursor coding platform — already running the models inside their products. The joint report from House Homeland Security Committee Chairman Andrew Garbarino and House Select Committee on China Chairman John Moolenaar makes 12 recommendations and warns that Beijing is closing the artificial intelligence gap with American technology that was, in part, taken from American systems.
The report, released on Friday, November 20, 2026, concludes a joint investigation launched in April by the two committees. It examined where Chinese-developed open-weight models are used in the United States, how they were trained, and whether their capabilities were advanced by unauthorised copying — known as distillation — from leading American systems.
At the centre of the report are three American companies. Airbnb uses models developed by the Chinese technology giant Alibaba in customer support. Anysphere, the company behind the Cursor coding platform, built its flagship proprietary model on a base model from Moonshot AI, a Chinese developer. DoorDash told the committees that multiple families of Chinese-developed models had been authorised and deployed across its business. The chairmen call for prohibiting Chinese-developed models in federal systems and federal contract work, removing them from high-risk functions of critical infrastructure, and barring federal cybersecurity grant funds from being used to buy them.
The report lands in the middle of a widening Washington fight over Chinese artificial intelligence. Earlier this week, Reuters reported that the Trump administration is weighing a plan to require US companies to publicly disclose their use of Chinese AI models — a "name and shame" approach that would expose how deeply the models have penetrated American business. It also sits alongside the Nvidia espionage scandal around former Megaspeed executive Alice Huang. Top House Republicans had already asked Nvidia chief executive Jensen Huang to brief Congress behind closed doors on the affair.
In September 2026, the National Security Agency, the Cybersecurity and Infrastructure Security Agency (CISA) and the FBI issued a joint advisory warning about the national security impact of large-scale distillation efforts against American AI companies, saying China uses advanced AI models to increase the scale, scope and speed of its intrusions into US systems. China, for its part, is accelerating its own comprehensive AI legislation.
The report's central finding is that Chinese-developed open-weight models are already widely used across the American economy, but much of that use is hard to track. Cost, customisation and near-frontier performance are driving adoption, and the United States lacks a competitive open-weight alternative at that price and performance. Federal investigations have found that much of the Chinese progress rests on capabilities extracted from American models.
The chairmen warn the risks are growing as the Chinese models become more capable, while their behaviour and security properties can differ in ways ordinary evaluations may not detect. Beijing is using open-weight distribution, state support and low prices to expand global dependence on its models. That is why the chairmen say the moment to act is now, before the models are embedded so deeply that removing them would require another costly rip-and-replace effort.
In the foreword to the report, Garbarino and Moolenaar wrote: "The investigation uncovered risks that American businesses, policymakers, and national security officials can no longer afford to overlook or downplay. ... While America is investing billions and racing to build 'Super Intelligence,' Beijing is using stolen intelligence to close the gap." They added: "We should not wait five years to discover that much of the world, and much of the United States, runs on AI shaped by the CCP. ... The time to act is now, while we still have the opportunity and leverage to do so."
The companies responded differently. Anysphere told the committees it plans to explore phasing out Chinese-developed open-weight models because of geopolitical risk. DoorDash did not change its use of a Moonshot model after the White House accused Moonshot of illicit distillation. Airbnb kept Alibaba models approved after the Department of War identified Alibaba as a Chinese military company.
Federal systems and critical infrastructure are the backbone of American government and daily life — power grids, water systems, hospitals, transport and financial networks. If the artificial intelligence running inside them was built in China, trained in part on copied American technology, and shaped by rules set in Beijing, Washington's leverage over its own systems weakens. The report argues that the danger is not only espionage but dependence: once a country's infrastructure runs on another country's models, switching away becomes expensive and disruptive. And if American companies keep choosing cheap Chinese models, the report warns, the country could lose the market that decides whose technology becomes the global default.
The recommendations are not yet law. Turning them into rules would require legislation, executive action, or both. The committees will continue their oversight, and the companies named in the report now face a choice: phase out the Chinese models voluntarily, as Anysphere says it will explore, or wait for Washington to force the issue.
What are open-weight AI models?
AI systems whose trained weights are published, so anyone can download, customise and run them. They spread quickly — and are hard to track once deployed.
Which companies did the investigation examine?
Airbnb, which uses Alibaba models in customer support; Anysphere, whose Cursor coding platform was built on a Moonshot AI base model; and DoorDash, which reported multiple Chinese model families authorised and deployed.
What did the report recommend?
Twelve recommendations, including prohibiting Chinese-developed models in federal systems and federal contract work, removing them from high-risk critical infrastructure functions, and launching a national American open-weight AI initiative.
Are Chinese AI models banned from federal systems now?
Not yet. The recommendations have not been enacted into law. They build on a Department of Homeland Security order excluding DeepSeek from federal civilian agency systems.
What is the proposed American open-weight AI initiative?
A national effort to build competitive American open-weight models so US companies and cyber defenders are not forced to choose between expensive proprietary American models and cheap Chinese alternatives.