House Homeland Security and China Select committee chairmen released a joint report, The Record reported1 Nov. 20, recommending that China-developed AI models be barred from federal contract work and high-risk critical-infrastructure functions.

The 12 recommendations are congressional proposals, not enacted law, and are narrower than a general ban on private-sector use. The committees' announcement2 also lists a recommendation to "govern both closed and open models and avoid overregulation." The committees' page is undated and says only that the report was released "today." The report itself was not reviewed for this article. Its contents are described here as given in that announcement and in The Record's account, which puts the report at 50 pages.

Background: letters, warnings and a Huawei comparison

The committees began the investigation in April 2026 with letters to Anysphere and Airbnb. A letter to DoorDash followed in July 2026. A July 31 committee release3 said the committees had held "productive discussions" with Anysphere and Airbnb. The committees' November announcement says both companies have engaged with them on the technical, commercial and security considerations of using Chinese open-weight models.

The July 31 release also quoted the committees' letter to DoorDash, which acknowledged that companies adopt such models for competitive capabilities, lower costs and greater customization. The letter said those considerations "do not eliminate the need for risk-based safeguards."

The release cited a July 22 statement by the Director of the White House Office of Science and Technology Policy. According to the release, the director said Chinese AI company Moonshot may have operated a covert platform for large-scale distillation against American AI models. The director also said Moonshot trained models using advanced systems it was not authorized to obtain. In September 2026, the committees say, the NSA, CISA and the FBI issued a joint advisory on the national security impact of large-scale distillation against American AI companies.

The Record reported that the report points to Huawei and ZTE equipment, which quickly found a place in the U.S. market and later threatened American networks. By that comparison, the report warns, removing China-developed open-weight models after they take root would take several years and cost taxpayers billions.

What the report says about the companies and the models

According to The Record, the report focuses on Airbnb, DoorDash and Anysphere, owner of the Cursor coding platform, as examples of companies deploying Chinese open-weight models. The Record said the companies used them for customer assistance, coding projects and taking inventory. Airbnb, DoorDash and Anysphere did not immediately respond to The Record's requests for comment.

The Record said the authors, Reps. Andrew Garbarino and John Moolenaar, drew on hearings, classified briefings and agency advisories. They also received briefings from the three companies. The report says U.S. businesses are adopting Chinese models because they are inexpensive, customizable and competitive with leading American systems. It adds that federal investigations found major Chinese developers "used proxy networks and fraudulent accounts to harvest outputs from American frontier models and replicate their capabilities."

The committees also describe links between Chinese AI developers and the military. They say personnel associated with the 1st Bureau of the People's Liberation Army Cyberspace Force, which they describe as the unit responsible for cyber espionage and attacks, were present on multiple occasions at Zhipu's Beijing headquarters, most recently in August 2026. The committees characterize those visits as indicating a channel between the PLA's cyber intelligence apparatus and one of China's leading AI developers.

The committees say their review of PLA procurement records found the following:

  • A July 2026 procurement for a large PLA cyber range required integrating DeepSeek to support "intelligent attack" and "intelligent penetration."
  • Other records show a command information system built around Alibaba's Qwen and DeepSeek.
  • A 2026 project required models from DeepSeek, Alibaba, Moonshot and MiniMax to run on a network cut off from the internet for military AI agent development.

The committees add that documents produced to them show these same model families running inside major American companies.

The recommendations

The 12 recommendations include:

  • Launching a national American open-weight AI initiative.
  • Establishing a model provenance standard and an AI bill of materials.
  • Evaluating Chinese models systematically.
  • Making CISA the single point of submission for AI-related cyber incident reports.
  • Examining sales of American training data to Chinese AI laboratories.
  • Extending federal restrictions "beyond DeepSeek" and keeping China-developed models out of federal contract work and high-risk critical-infrastructure functions. The committees' page does not describe the restrictions already in place.

Mark Montgomery, a cybersecurity expert at the Foundation for Defense of Democracies, told The Record the report matters because it identifies the risks of the U.S. not competing seriously in the low-cost open-source model market. He said China, more limited at the frontier, has pursued a strategy to expand global dependence on its open-source AI. "Companies and utilities, driven by bottom line resource decisions will, as always, ignore security and safety issues in pursuit of a lower front end cost," he said. "China knows this recipe well."