# US Moves to Stop Chinese AI From Becoming the "Next Huawei"

Canonical: https://ournews.ai/news/beijings-ai-playbook-six-takeaways?lang=en
Published: 2026-11-20T19:03:54.180643+00:00
Updated: 2026-11-21T04:51:27.620381+00:00

*Cover: China and the US competing for AI dominance/Ournews.ai*

Two House committees released a seven-month investigation on November 20 into how **Airbnb, Cursor and other US companies use Chinese AI models**. Their report, [**Beijing’s AI Playbook**](https://homeland.house.gov/wp-content/uploads/2026/11/2026.11.20-Joint-Report-PRC-AI-Models.pdf), calls for action to avoid the “next Huawei”: dependence on Chinese technology that becomes expensive to replace. Here are its six main takeaways.

**1. Chinese models already handle everyday work for US companies.**

Airbnb uses Alibaba’s Qwen models for an assistant handling about **80,000 guest and host interactions a day**, plus summaries of roughly **250,000 support cases and calls daily**. Cursor’s maker, Anysphere, built its Composer 2 coding model on Moonshot’s **Kimi K2.5**. These are tools inside American products, even when customers never see the Chinese developer’s name. [Company records, pages 6 and 13](https://homeland.house.gov/wp-content/uploads/2026/11/2026.11.20-Joint-Report-PRC-AI-Models.pdf#page=7)

**2. Low prices and customization make them attractive.**

Companies can download open-weight models, adapt them and run them on infrastructure they choose. The committees say US alternatives struggle to match the combination of price, capability and control. A company can therefore choose Chinese AI for practical business reasons without sending its customer requests to a server in China. [Pages 16–19](https://homeland.house.gov/wp-content/uploads/2026/11/2026.11.20-Joint-Report-PRC-AI-Models.pdf#page=17)

**3. The committees allege some capabilities were extracted from US models.**

The report cites federal assessments of fraudulent accounts and proxy networks used to collect American models’ outputs for training. That practice, known as distillation, can help another model imitate the system producing the answers. The report calls for action against illicit extraction and scrutiny of American training-data sales to Chinese labs. [Pages 20–22 and recommendations](https://homeland.house.gov/wp-content/uploads/2026/11/2026.11.20-Joint-Report-PRC-AI-Models.pdf#page=21)

**4. Replacing an embedded model can take months.**

DoorDash reported **25 Chinese model families authorized and deployed**. It said replacing one model embedded in a larger system could take **up to six months**, although most other uses could be switched almost immediately. That is the practical concern behind the Huawei comparison: switching costs grow after software becomes part of daily operations. [Pages 6 and 27–28](https://homeland.house.gov/wp-content/uploads/2026/11/2026.11.20-Joint-Report-PRC-AI-Models.pdf#page=7)

**5. US hosting does not settle every security question.**

Running a model locally reduces direct exposure of data to its developer. The committees also raise concerns about the model itself, citing tests involving insecure code under particular prompts and safeguards that can be removed. Their argument is that companies need to examine both where data goes and how the model behaves. [Pages 29–33](https://homeland.house.gov/wp-content/uploads/2026/11/2026.11.20-Joint-Report-PRC-AI-Models.pdf#page=30)

**6. The report proposes restrictions alongside American alternatives.**

The chairmen want Chinese models excluded from federal systems and contract work, removed from high-risk functions in critical infrastructure, and subject to clearer origin disclosures. They also call for a national American open-weight AI initiative. **These are the report’s recommendations, not a new blanket ban on private businesses using Chinese AI.** [Recommendations, pages 35–38](https://homeland.house.gov/wp-content/uploads/2026/11/2026.11.20-Joint-Report-PRC-AI-Models.pdf#page=36)

The same competition runs through hardware: [what US records reveal about Chinese requests for Nvidia and AMD chips](https://ournews.ai/news/china-ai-chip-approvals-nvidia-amd).


Source: https://homeland.house.gov/wp-content/uploads/2026/11/2026.11.20-Joint-Report-PRC-AI-Models.pdf
