# Congress Examines Security Risks as Businesses Use Lower-Cost Chinese AI > Source: > Published: 2026-07-10 19:08:54+00:00 TL;DR — Key Takeaways **Congress is scrutinizing U.S. companies that use lower-cost Chinese AI models, citing data security, censorship and national security risks.****Chinese AI models are gaining traction with U.S. businesses because they are cheaper, but lawmakers worry those savings may come with security and geopolitical risks.****As Chinese open-weight models win enterprise users on price and performance, Congress is weighing whether new restrictions or domestic incentives are needed.** Congress is stepping up scrutiny of American companies that use AI models developed in China, voicing concerns that lower-cost Chinese models could create national security and economic risks. The House Committee on Homeland Security and the House Select Committee on the Chinese Communist Party launched a joint investigation in April into the use of Chinese AI models by US businesses. The concern is that Chinese developers continue to narrow the performance gap with leading American models while offering substantially lower operating costs. The investigation has focused initially on AI coding startup Cursor and rental company Airbnb. Lawmakers requested information about each company’s use of Chinese-developed models and the safeguards in place to protect sensitive data. Andrew Garbarino, chairman of the House Committee on Homeland Security, said the rapid improvement of Chinese AI capabilities presents a growing concern for cybersecurity. He pointed to reports showing that some Chinese open-weight models can perform vulnerability discovery and cybersecurity tasks at levels comparable with leading US systems. The State Department also expressed concern over the trend, saying that Chinese AI models reflect the priorities and political objectives of Beijing, including censorship and ideological controls. Chinese officials rejected those accusations, saying criticism of the country’s AI industry is unfounded and politically motivated. **A Shift in Enterprise AI Purchasing** Over the last year, there has been a shift in enterprise AI purchasing. While several US government agencies have prohibited the use of Chinese AI services such as DeepSeek, private companies are free to use them. Lower costs have made those models attractive to startups and software developers looking for lower AI costs. AI startup Lindy has publicly stated that moving workloads to China’s DeepSeek model is expected to save the company millions of dollars. Coinbase CEO Brian Armstrong has also spoken favorably about using Chinese AI models when they provide better value. Cursor’s Composer 2 coding model was built using Kimi, an AI model developed by Chinese startup Moonshot AI. Airbnb, meanwhile, said the vast majority of its AI activity relies on US-developed models, adding that its limited use of Chinese open-source models is handled through approved US-based providers that separate customer data from the underlying models. Industry data indicates that adoption is accelerating. According to reports cited by lawmakers, Chinese open-weight models accounted for more than 30% of weekly token usage by US companies through OpenRouter after February 8, reaching a peak of 46%. That compares with an average of 11% during the previous year. Pricing is a key driver. Reports indicate that several leading Chinese open-weight models cost 60% to 90% less than comparable offerings from OpenAI and Anthropic, creating a strong financial incentive for organizations managing growing AI workloads. Congress is also examining whether the US has provided enough competitive alternatives. Committee staff are evaluating whether American developers have sufficient access to affordable open-weight models or whether businesses are being pushed toward Chinese software because domestic options remain more expensive or carry licensing restrictions. Lawmakers acknowledge that restricting Chinese AI will be difficult because many of the models are open-weight, allowing users to download and run them independently. Freely available models make outright bans challenging to enforce and could complicate any regulatory approach. The investigation overlaps with concerns raised by leading US AI companies over model distillation, a technique that uses the outputs of existing AI systems to train new models. American developers have argued that Chinese competitors have used distillation to accelerate development of advanced capabilities. The committee has not proposed specific legislation, but lawmakers say the goal is to determine whether new procurement rules or incentives for domestic AI development are needed. Going forward, the US government faces the challenge of balancing national security concerns with businesses’ growing adoption of foreign lower-cost AI models.