politics

Cursor, a subsidiary of SpaceX, plans to gradually phase out Chinese AI models.

Published November 20, 2026 · 3 min read
A Kimi logo is pictured at the Moonshot AI stand, featuring the Kimi K3 model, during the World Artificial Intelligence Conference (WAIC) in Shanghai on July 18, 2026. (Photo by HECTOR RETAMAL / AFP via Getty Images)

According to recent reports from financial news outlet CNBC, SpaceX’s artificial intelligence programming platform, Cursor, is contemplating the ceasing of its use of Chinese AI models due to “geopolitical risks.”

In August of this year, SpaceX acquired Cursor at an implied valuation of $60 billion. Cursor had previously developed its programming model, “Composer 2,” based on the Kimi K2.5 model created by the Chinese AI startup, Moonshot AI.

The U.S. Congress initiated a joint inquiry among committees in April to investigate the adoption of Chinese open-weight AI models, spurred by the alarming rate of their integration into American enterprises. Lawmakers expressed concerns that even if companies download the models for local or U.S. cloud operation, they face significant risks related to intellectual property disputes, cybersecurity vulnerabilities, algorithm scrutiny, and potential infiltration of critical infrastructure.

A report jointly released by the House Committee on Homeland Security and the House Select Committee on the Chinese Communist Party indicated that prominent U.S. consumer and developer platforms, such as Cursor (Anysphere), Airbnb, and DoorDash, have integrated Chinese open-weight models into their internal operations or product offerings, prompting Congressional scrutiny.

This report is the culmination of several months of investigatory efforts aimed at understanding the role of Chinese AI models within the U.S. tech ecosystem. It reveals that Cursor has informed the investigative committees about its plans to explore a phased elimination of models developed by the People’s Republic of China (PRC) deemed to pose geopolitical risks.

Moreover, Airbnb and DoorDash have also participated in the inquiry, providing details regarding their use of Chinese AI models.

Requests for comments from Cursor, SpaceX, and DoorDash went unanswered, while Airbnb declined to provide any remarks. The U.S. Department of Commerce’s Bureau of Industry and Security (BIS) did not respond immediately to an inquiry.

Data from the Congressional committee report illuminates a rapid increase in the share of AI workloads attributable to Chinese-developed models. This proportion surged from approximately 1% at the end of 2024 to around 30% by the end of 2025. The report attributes this growth to the low deployment costs and high customizability of models developed by companies such as DeepSeek, Moonshot AI, Alibaba (Qwen), and MiniMax. Meanwhile, the U.S. is noted to lack a competitive open-weight model alternative that meets the required market price and performance levels.

The findings indicate that Beijing is actively seeking to translate these technological advancements into significant global market shares while establishing long-term dependency relationships.

Further, the report proposes several recommendations to address the highlighted concerns. These include launching a “national-level open-weight program” in the United States to consolidate public computing resources, share assessment infrastructures, and allow access to frontier American AI models for training, provided it is done legally and in compliance with regulations.

Additionally, the report suggests further expansion of an existing ban issued in October that prevents DeepSeek from entering U.S. federal civil agency systems, explicitly prohibiting the use of Chinese-developed AI models within federal information systems and during the execution of federal contracts.

The U.S. also expressed apprehensions regarding the ability of Chinese enterprises to remotely access Nvidia chips through overseas data centers and their capacity to enhance their capabilities by employing “model distillation” technology, which utilizes older, more established models to assist in training newer models.

In a related move, the BIS announced a temporary final rule on Monday aimed at curbing the ability of Chinese enterprises to remotely acquire AI chips, marking a significant step in restricting foreign influences on American technology ecosystems.