San Francisco-based DoorDash is facing a new Washington headache after House lawmakers demanded information about the company’s use of artificial intelligence models developed in China. The request puts a cost-cutting engineering decision at the center of a much bigger fight over national security, software supply chains, and America’s ability to compete in open AI.
The House Committee on Homeland Security and the House Select Committee on the Chinese Communist Party sent DoorDash a letter requesting information and documents about its evaluation and deployment of China-developed AI systems, according to CNBC. A committee aide said the panels are also examining whether the United States has a sufficient strategy for open-weight AI models, which can be downloaded, modified, and hosted by companies themselves.
The request is part of a broader joint investigation led by Chairmen Andrew Garbarino and John Moolenaar into the security implications of Chinese AI models. In an announcement from the House committees, lawmakers said the inquiry also involves models connected to companies such as DeepSeek, Alibaba, Moonshot AI, and MiniMax, with separate scrutiny directed at Airbnb and Anysphere, the company behind Cursor.
DoorDash’s Cost-Saving AI Mix Put It In The Spotlight
DoorDash does not appear to have abandoned American AI providers altogether. Instead, company co-founder and chief technology officer Andy Fang has described a hybrid approach in which the company sends lower-level work to Kimi K2.6, developed by Beijing-based Moonshot AI, while reserving Anthropic’s Fable 5 for more difficult tasks.
Fortune reported that Fang characterized the combination as delivering better quality at a cheaper cost than DoorDash’s previous setup using Anthropic models. DoorDash’s AI research team has been testing these systems for software-development work, making the company’s model selection a practical engineering decision rather than a purely theoretical experiment.
The local angle is hard to miss. DoorDash is headquartered in San Francisco, according to its SEC filing, so a national debate over Chinese technology is now landing on one of the city’s most prominent tech companies and its AI research operation.
Why Cheaper Chinese Models Are Gaining Ground
The economics are helping drive the trend. AP News reported this week that Chinese AI models are gaining traction in the United States because they are often cheaper, openly distributed, and increasingly competitive with premium American systems.
That creates an awkward calculation for companies like DoorDash: use a high-priced frontier model for every task, or route routine workloads through a less expensive open-weight system and keep the most powerful tools for jobs that genuinely need them. The savings can be meaningful at scale, but the lawmakers’ concern is that price and performance may not capture risks involving data handling, model behavior, or dependence on technology developed by Chinese companies.
No Blanket Ban Yet, But Scrutiny Is Tightening
U.S. companies are not currently prohibited from adopting Chinese AI models, and DoorDash said it supports American AI leadership while working to ensure AI is used safely and responsibly, according to CNBC. The company also said it would engage with the committees about its use of American frontier models and open-weight systems.
The House committees have argued that practical advantages such as lower cost, customization, and an alternative to proprietary providers do not erase national-security concerns. Their investigation announcement says the federal government should take a risk-based approach while strengthening American open-weight alternatives, a position that could eventually influence procurement rules, corporate guidance, or restrictions on certain systems.
For now, DoorDash has been asked to explain how it evaluates and deploys these models. The larger question for San Francisco’s tech industry is whether a cheaper AI tool remains a bargain once lawmakers begin pricing in cybersecurity, supply-chain, and geopolitical risk.
