November 21, 2026

Congress Warns American Companies Off Chinese Open-Weight AI, Naming Airbnb, DoorDash and the Maker of Cursor

A joint report from two House committees calls the spread of China-developed open-weight AI models inside American tech stacks alarming, compares it to the rise of Huawei and ZTE equipment, and offers 12 recommendations, including an AI bill of materials. Here is the story behind the report and what it means for businesses using AI tools.

A young software engineer in a charcoal hoodie working on a laptop in a bright office

Open-weight AI models, whose trained parameters anyone can download and run, have become a popular way for companies to add AI features without paying per question to a big provider. A new congressional report argues that some of the most popular of those models come from China, and that American companies building on them are repeating a mistake the country has already paid for once. The account below is based on The Record’s coverage of the report.

Two committees, one warning

The 50-page report was produced jointly by the House Homeland Security Committee and the Select Committee on the Chinese Communist Party, and co-authored by the committees’ chairmen, Representatives Andrew Garbarino and John Moolenaar. It draws on hearings, classified briefings and agency advisories, as well as briefings the lawmakers received from three companies about how they had built China-developed open-weight models into their consumer platforms, enterprise software and other systems.

Those three companies, Airbnb, DoorDash and Anysphere, the company behind the Cursor coding platform, serve as the report’s examples. According to the report, the models were being used for customer assistance, coding work and taking inventory. None of the companies immediately responded to The Record’s request for comment.

The Huawei comparison

The report’s central argument is historical. Lawmakers point to how quickly telecom equipment from Huawei and ZTE spread through the market before it was judged a threat to American networks, and warn that if Chinese open-weight models are allowed to take root in the same way, pulling them out later could take several years and cost taxpayers billions.

The report does not pretend the appeal is mysterious. It says startups and Fortune 500 companies alike are choosing these models because they are inexpensive, easy to customize and perform well. But it also cites federal investigators’ findings that Chinese developers used proxy networks and fraudulent accounts to collect outputs from American frontier models in order to copy their capabilities. The report frames the choice bluntly, saying the risks it uncovered can no longer be overlooked or played down.

Twelve recommendations

Among the report’s 12 policy proposals:

  • An American open-weight AI initiative to give companies a domestic alternative.
  • A standard for documenting where models come from, and an AI bill of materials listing what is inside an AI system.
  • Barring China-developed models from federal contract work and from sensitive critical infrastructure functions.
  • Systematic evaluation of Chinese models.

An outside view

Mark Montgomery, a cybersecurity expert at the Foundation for Defense of Democracies, told The Record the report matters because it explains the risk of the U.S. not competing seriously in the market for low-cost open models. In his view, China has pursued global dependence on its open-source AI because it trails in frontier models, and businesses focused on up-front cost tend to overlook security until it is too late.

What this means for your company

The report is a set of recommendations, not a law, and it does not ban anyone from using any model today. But it is a clear signal of where policy may be heading, especially for companies that do government contract work or serve critical infrastructure. Even if neither applies to you, it is a good prompt to find out which AI models are actually running in your business. Many software products and coding tools use third-party models behind the scenes, and staff may be trying free or downloadable ones on their own. Ask your key software vendors which models their AI features rely on and where your data goes. Write a short AI use policy that says which tools are approved and what information must never be pasted into them. And if you contract with government agencies, start keeping a simple record of the AI components in what you deliver, since documentation requirements tend to arrive faster than expected.

If you would like a clear picture of which AI tools and models your team is using and whether that fits your contracts and risk tolerance, we can help you build it. Reach out to Layer 7 Data Solutions to discuss how we’ve helped companies similar to yours.

About The Author