Threats

House Committees Warn US Firms Against Embedding Chinese Open-Weight AI Models

November 21, 2026 08:00 · 8 min read
House Committees Warn US Firms Against Embedding Chinese Open-Weight AI Models

Congressional Investigation Highlights Risks of Chinese AI Adoption

A joint investigative report from the House Homeland Security Committee and the Select Committee on the Chinese Communist Party warns that U.S. companies integrating China-developed open-weight artificial intelligence models into their technology stacks face significant national security risks. The 50-page document, co-authored by Representatives Andrew Garbarino (R-NY) and John Moolenaar (R-NY), draws on hearings, classified briefings, agency advisories, and direct briefings from the companies involved.

The investigation specifically identifies Airbnb, DoorDash, and Anysphere — owner of the Cursor coding platform — as examples of firms deploying Chinese open-weight models for customer assistance, coding projects, and inventory management. None of the three companies responded immediately to requests for comment.

Open-Weight Models and the Strategic Threat

Open-weight models publish their numerical weights and biases for public download, making them easily accessible and customizable. The report cautions that allowing these models to take root in U.S. infrastructure mirrors the earlier proliferation of Huawei and ZTE technology, which threatened American networks and would require years and billions of taxpayer dollars to remove.

"The investigation uncovered risks that American businesses, policymakers, and national security officials can no longer afford to overlook or downplay," the report states. "While America is investing billions and racing to build 'Super Intelligence,' Beijing is using stolen intelligence to close the gap."

How Chinese Models Gain Market Traction

According to the findings, Silicon Valley startups and Fortune 500 companies alike are adopting Chinese open-weight models because they are inexpensive, easily customized, and perform well. However, federal investigators found that Chinese developers "used proxy networks and fraudulent accounts to harvest outputs from American frontier models and replicate their capabilities."

Mark Montgomery, a cybersecurity expert at the Foundation for Defense of Democracies, said the report is important because it identifies the risks posed by the U.S. not competing seriously in the low-cost open-source model marketplace. He noted that China, more limited in the frontier model space dominated by U.S. firms such as Anthropic and OpenAI, has pursued a strategy to expand global dependence on its open-source AI using the "traditional Chinese model of state policy and capital" to kill competition.

"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," Montgomery said. "China knows this recipe well."

Twelve Policy Recommendations for Lawmakers

The report outlines 12 policy recommendations designed to counter the threat:

The remaining recommendations address supply chain transparency, export controls, intelligence sharing, and workforce development to ensure the United States maintains a competitive edge in open-weight AI without relying on adversarial technology.

Long-Term Implications for Critical Infrastructure

The lawmakers emphasize that the cost of inaction extends beyond immediate security vulnerabilities. If Chinese open-weight models become embedded across consumer platforms, enterprise software, and critical infrastructure, the eventual remediation effort would parallel the multi-year, multi-billion-dollar campaigns required to purge Huawei and ZTE equipment from U.S. telecommunications networks.

The report's findings reflect a broader congressional push to secure the AI supply chain and ensure that American innovation — not subsidized foreign alternatives — powers the next generation of critical systems.


Source: The Record

Source: The Record

Powered by ZeroBot

Protect your website from bots, scrapers, and automated threats.

Try ZeroBot Free