# House chairs warn of U.S. reliance on Chinese AI models

Published 2026-11-21 · AI.info Team · AI.info

> A joint House investigation found Chinese-developed AI models in use at major American companies and urged restrictions in sensitive systems. The committee chairs also called for a national initiative to develop U.S. open-weight alternatives.

American companies are increasingly using Chinese-developed open-weight AI models, prompting two [House](https://chinaselectcommittee.house.gov/media/press-releases/joint-investigation-report-china-is-trying-to-get-us-companies-addicted-to-chinese-ai-models) committee chairs to call for restrictions on their use in federal work and sensitive infrastructure. The joint investigation found that cost, customization and performance were driving adoption, while much of the use was difficult to track.

The report was released on November 20 by John Moolenaar, the Republican chairman of the House Select Committee on China, and Andrew R. Garbarino, the Republican chairman of the House Committee on Homeland Security. The chairs called for Chinese-developed models to be barred from federal systems and contract work and removed from high-risk functions of critical infrastructure.

## Companies examined

The investigation began in April 2026 with letters to Anysphere and [Airbnb](https://ai.info/companies/airbnb) about their evaluation and use of Chinese open-weight models. In July, the chairs sought similar information from DoorDash.

Airbnb uses Alibaba models in customer support, while Anysphere built the flagship proprietary model in its [Cursor](https://ai.info/companies/anysphere-inc) coding platform on a [Moonshot AI](https://ai.info/companies/moonshot-ai) base model, according to the Homeland Security Committee. DoorDash reported that multiple Chinese model families had been authorized and deployed.

The committees said Chinese models were attracting users because they were inexpensive, customizable and competitive with leading American systems. Their findings also said the United States lacked an open-weight alternative that was competitive at the price and performance the market demanded.

The inquiry examined where the models came from, the data and techniques used to train them, and how they were made available to companies. It also considered whether the U.S. open-weight strategy offered companies an alternative to choosing between expensive or proprietary American models and lower-cost Chinese ones.

## Security and training concerns

The report also examined whether Chinese developers advanced their models through unauthorized distillation from leading U.S. systems. The chairs said federal investigations had found that major Chinese developers used proxy networks and fraudulent accounts to collect outputs from American frontier models and replicate their capabilities.

The investigation also found personnel associated with the 1st Bureau of the People’s Liberation Army Cyberspace Force at Zhipu’s Beijing headquarters on multiple occasions, most recently in August 2026. The committees described the visits as indicating a channel between the military’s cyber intelligence apparatus and the AI developer.

The committees said a July 2026 procurement for a large military cyber range required DeepSeek integration to support capabilities described as intelligent attack and intelligent penetration. Other procurement records included a 2026 project requiring models from DeepSeek, Alibaba, Moonshot and [MiniMax](https://ai.info/companies/minimax) to run on a network cut off from the internet for military AI agent development.

Documents provided to the committees showed that the same model families were running inside major American companies, the chairs said. They also warned that model behavior and security properties could differ in ways ordinary evaluations might not detect.

## Proposed response

Among 12 recommendations, the chairs called for a national American open-weight AI initiative, a model provenance standard and an AI bill of materials. They also proposed barring federal cybersecurity grant funds from being used to buy Chinese-developed models.

Other recommendations include systematically evaluating Chinese models, creating a developer-level AI threat channel and examining sales of American training data to Chinese AI laboratories. The chairs also proposed making the Cybersecurity and Infrastructure Security Agency the single submission point for AI-related cyber incident reports and establishing an independent board to investigate significant incidents.

## Sources

1. Original source · Government or agency page [House Select Committee on China](https://chinaselectcommittee.house.gov/media/press-releases/joint-investigation-report-china-is-trying-to-get-us-companies-addicted-to-chinese-ai-models) chinaselectcommittee.house.gov

Other pages read

2. [House Committee on Homeland Security](https://homeland.house.gov/2026/11/20/chairmen-garbarino-moolenaar-release-joint-investigative-report-urging-removal-of-prc-open-weight-models-from-federal-systems-high-risk-critical-infrastructure/) homeland.house.gov

## Related stories

- [Comer Seeks Records on Delaware’s AI-Run Company Plan](https://ai.info/news/comer-seeks-records-delaware-ai-run-company-plan)
- [Thinking Machines releases Inkling 2 Mini, an open-weight model small enough for a laptop](https://ai.info/news/thinking-machines-inkling-2-mini-open-weights)
- [China Sets 18 Tasks to Cushion AI’s Impact on Jobs](https://ai.info/news/china-ai-employment-guidelines-job-displacement)
- [Treasury Proposes Broader U.S. Investment Ban on Chinese AI Firms](https://ai.info/news/treasury-proposes-broader-us-investment-ban-chinese-ai-firms)
- [Researcher says six Chinese AI models carry Blackwell chip traces](https://ai.info/news/chinese-ai-models-blackwell-chip-fingerprints)

## Explore

[More articles](https://ai.info/news)

---

Canonical: https://ai.info/news/house-investigation-chinese-ai-model-dependence
