US House committees urge limits on Chinese superintelligence models

Their investigation identifies uses at Airbnb, DoorDash and Cursor maker Anysphere, and recommends restrictions for sensitive systems; the proposals are not adopted rules.
In this article
The chairs of two US House committees released an investigation on 20 November identifying how Airbnb, DoorDash and Anysphere use China-developed open-weight superintelligence (SI) models. They recommend restricting such models in federal systems, government contract work and high-risk critical infrastructure functions.
The proposals are not binding rules. They would go beyond an existing Department of Homeland Security order excluding DeepSeek and affiliate products and services from federal civilian purchases and requiring their removal from agency systems, including third-party products developed using DeepSeek implementations.
Open-weight models make numerical parameters used by the model available for others to download and adapt. The report says companies choose Chinese models because they are inexpensive, customizable and competitive with leading US systems.
How the companies use the models
Airbnb runs customized Qwen models from Chinese developer Alibaba on US-based infrastructure, including a customer-support assistant handling about 80,000 guest and host interactions a day, the report says. Airbnb separately said its SI activity runs overwhelmingly on US-origin models and that its limited Chinese-origin models run through approved US-based providers, keeping data and operations separate and protected.
DoorDash told the committees it had authorized and deployed 25 Chinese model families in a non-trivial way. Its co-founder said the company routes some internal tasks through Kimi, made by Moonshot AI. DoorDash told the committees that replacing a model embedded in a larger system could take up to six months.
Anysphere built Composer 2, the flagship proprietary model in its Cursor coding platform, on Moonshot AI’s Kimi K2.5, the report says. Anysphere told the committees it plans to explore phasing out Chinese-developed open-weight models because of geopolitical risk; it did not commit to completing a phaseout.
The investigation and its concerns
Republican representatives Andrew Garbarino, who chairs the House Homeland Security Committee, and John Moolenaar, who chairs the House select committee on China, began the inquiry with letters to Airbnb and Anysphere on 29 April. They sought similar information from DoorDash in late July and reviewed company responses and briefings as well as classified briefings.
The chairs argue that reliance on the models creates national-security, cybersecurity and economic risks, including dependence and costly replacement later. The report cites federal investigations that found major Chinese developers used proxy networks and fraudulent accounts to collect outputs from leading US models and replicate their capabilities.
That process, called distillation, can be a legitimate way to train one model using another’s outputs. The chairs distinguish it from methods involving fraudulent accounts, proxy networks or evasion of access restrictions. China’s embassy in Washington told Reuters that Beijing opposes politicizing trade and technology, saying it would stifle global SI advances and that China would safeguard its businesses' lawful rights.
What the chairs recommend
The report’s 12 recommendations include a national US open-weight SI initiative, a standard for disclosing where models come from and an SI bill of materials identifying their components. The chairs also recommend barring federal cybersecurity grant funds from buying Chinese-developed models.
The other proposals call for promoting US models internationally; extending threat-information sharing and creating a developer threat channel; evaluating Chinese models systematically; and preparing federal agencies to defend networks and secure SI agents. The chairs also recommend making the Cybersecurity and Infrastructure Security Agency the central point for SI cyber-incident reports and creating an independent board to investigate major incidents.
Further recommendations address illicit distillation and the gray market, sales of US training data to Chinese laboratories, governance of both closed and open models, and cooperation with allies. The committees said they would continue oversight.
Separately, Reuters said three people told it the Trump administration was considering requiring US companies to publicly disclose their use of Chinese models. It was unclear whether the idea would proceed to formal rulemaking.
Follow AveNews
Earlier in Technology
SoftwareWhatsApp patches Android flaw that could expose users’ IP information
WhatsApp’s security team found the flaw while disrupting a scam campaign. The company suspects a scammer used it to make a message seem more credible and has fixed two other security flaws.
3 min
GovernmentUS communications regulator allows prisons to jam contraband cellphones
The optional program requires licenses and testing. Wireless-industry association CTIA says a pilot blocked 911 calls and affected network equipment up to 2.5 miles away.
3 min
SuperintelligenceOpenAI says it strengthened reasoning safeguards in Astra and Sol
The company says internal tests found extraction success fell from 54.0% to 0.2% for earlier requests and from 22.9% to 0.9% for current requests, comparing different Astra generations and safeguards.
3 min
Sources
Show all 8 sourcesShow fewer
- homeland.house.govChairmen Garbarino, Moolenaar Release Joint Investigative Report Urging Removal of PRC Open-Weight Models From Federal Systems, High-Risk Critical Infrastructure – Committee on Homeland Security (opens in a new tab)
- news.sulat.comCursor may phase out Chinese AI models, House report says - Sulat AI & Tech News (opens in a new tab)
- govinfo.govFederal Register, Volume 91 Issue 201 (Tuesday, October 20, 2026) (opens in a new tab)
- The RecordUS companies’ use of Chinese open weight AI draws rebuke from House chairmen (opens in a new tab)
channelnewsasia.comTrump administration weighs forcing US firms to disclose Chinese AI use - CNA (opens in a new tab)
- homeland.house.govChairmen Garbarino, Moolenaar Announce Joint Investigation into National Security Risks Posed by PRC AI Models – Committee on Homeland Security (opens in a new tab)
- homeland.house.govChairmen Garbarino, Moolenaar Continue Joint Investigation Into Security Risks Posed by PRC Open-Weight AI Models – Committee on Homeland Security (opens in a new tab)
- nextgov.comIntelligence agencies warn of China’s large-scale AI model distillation efforts - Nextgov/FCW (opens in a new tab)
