The AI Race Between Washington And Beijing Continues. U.S. Lawmakers Are Scrutinizing The Use Of Chinese Models

As Artificial intelligence becomes another front in the contest between Washington and Beijing, U.S. lawmakers are examining the growing use of Chinese artificial intelligence models by American companies. CNBC reported that policymakers are weighing how to limit exposure to systems built by Chinese firms while businesses continue looking for cheaper tools.

Chinese models have gained ground in the United States because they have narrowed the performance gap with American systems while often costing far less to use. The issue has drawn attention from the House Committee on Homeland Security and the House Select Committee on China, which opened a joint investigation in April into the adoption of Chinese-developed models by U.S. companies.

The committees sent letters to Cursor and Airbnb as part of the probe, focusing on their use of or exposure to Chinese-built systems. Cursor built its Composer 2 model using Kimi, a model developed by Moonshot AI, while Airbnb told CNBC that its artificial intelligence work runs “overwhelmingly” on U.S.-origin models and that its limited use of China-origin models is handled through approved U.S.-based service providers.

The State Department told CNBC that the growing use of Chinese models by U.S. companies raises concerns because those systems are designed to advance Beijing’s narratives, censor dissent, and reflect Chinese Communist Party ideology. A spokesperson for the Chinese embassy in the U.K. rejected the criticism, telling the outlet that China opposes what it called baseless allegations against its technology sector.

The debate has intensified because adoption is no longer limited to testing. Chinese open models accounted for more than 30% of weekly token usage by U.S. companies through OpenRouter since Feb. 8, peaking at 46%, Resultsense reported. That was up from an 11% average over the previous year.

Cost is a central reason for the shift. Tekedia reported that leading Chinese open-source models are typically 60% to 90% cheaper than comparable offerings from OpenAI and Anthropic. The report said AI startup Lindy moved all of its traffic from Anthropic’s Claude to DeepSeek, a switch its CEO Flo Crivello said would save the company millions of dollars within months.

The growing use of Chinese models has also raised questions about data, model access, and the economics of American artificial intelligence companies. Business Insider reported that U.S. companies including Anthropic and OpenAI have raised concerns about distillation, a technique in which one model is trained using the outputs of another. Anthropic accused Alibaba of using unauthorized accounts to collect Claude outputs at scale, while OpenAI has warned that rivals could combine outputs from several U.S. models to replicate advanced capabilities.

The same issue was detailed by The New York Times, which reported that Anthropic sent a June 10 letter to Senators Tim Scott and Elizabeth Warren accusing Alibaba of copying its technology through distillation. The report said American companies argue that Chinese competitors have used the method to close the gap faster in areas such as business planning, drug research, surveillance, and military applications.

The national security concern is especially sharp in cybersecurity. Andrew Garbarino, chairman of the House Homeland Security Committee, told CNBC that recent reporting showing a Chinese open-weight model matching leading U.S. systems in some vulnerability discovery and cybersecurity tasks was “highly alarming.”

The Trump administration has already accused Chinese entities of running “industrial-scale campaigns” to copy U.S. artificial intelligence systems. At the same time, China is also considering restrictions on overseas access to its most advanced models, SAN reported, citing Reuters. Chinese officials have discussed limits on closed-source and open-source systems with companies including Alibaba, ByteDance, and Z.ai.

That has put U.S. companies in a difficult position. Many businesses are trying to control rising costs while avoiding security and compliance risks. Federal procurement restrictions are one option under discussion, CNBC reported, with experts saying the government could limit the use of Chinese models by agencies and contractors that work with Washington.

The challenge is that many Chinese models are open-weight, meaning their model files are already available online. Kyle Chan, a fellow at Brookings, told CNBC that banning them outright would be difficult because the files can be freely downloaded and shared. Daniel Remler, a senior fellow at the Center for a New American Security, told the outlet that restrictions could also affect startups that rely on lower-cost models.