House committees investigate Airbnb, Cursor-maker over Chinese AI model use

Two committees in the US House of Representatives have opened a joint investigation into American companies using Chinese-developed AI models, requesting details from CEOs about why their firms adopted the technology and what communications they've had with the Chinese model providers behind it.

Who's investigating, and who's named

The House Select Committee on China, chaired by John Moolenaar, and the House Homeland Security Committee, chaired by Andrew Garbarino, sent letters in late April to companies including Airbnb and Anysphere — the parent company of the AI coding tool Cursor. Airbnb has been using Alibaba's Qwen model in its customer-service tools, while Anysphere built its Composer 2 model on top of Kimi, developed by Beijing-based Moonshot AI.

The committees' letters requested details about each company's rationale for adopting Chinese AI systems and copies of communications with the model providers — a scope broad enough to cover both technical decision-making and any commercial relationship with the Chinese developers.

Why lawmakers are worried

The committees have framed their concern as a supply-chain and data-control risk, distinct from the more general AI safety debates that dominate coverage of the industry. Specific worries raised include content censorship built into the models' outputs, hidden vulnerabilities that could be exploited later, the risk of model distillation — where a Chinese provider could extract knowledge about how the US company operates from its own usage patterns — and exposure of sensitive operational data to a foreign-controlled model.

A State Department spokesperson went further in comments to CNBC, saying Chinese-designed models are built to "advance Beijing's narratives, censor dissent, and reflect CCP ideology and values" — a characterization that treats the models as a channel for state influence rather than a neutral software tool.

Why Chinese models are gaining ground anyway

The adoption trend the committees are responding to has a straightforward commercial driver: Chinese AI models have closed much of the performance gap with their American counterparts while remaining substantially cheaper to run. That combination has pulled in companies well beyond the two named in the congressional letters. Coinbase CEO Brian Armstrong has publicly described his company's use of two Chinese-made models, GLM 5.2 from Z.Ai and Kimi 2.7 from Moonshot AI, and reports have named Uber and the startup Lindy among firms exploring similar tools.

Airbnb has pushed back on the scope of the concern, telling the committees that most of its AI activity runs on US-origin models and that its limited use of Chinese open-source systems is routed through approved US-based service providers — an answer that concedes some Chinese-model usage while arguing the exposure is smaller and better controlled than the investigation implies.

Part of a broader confrontation

The investigation lands alongside separate friction over AI export policy on both sides. Reuters reported on July 7 that Chinese authorities are separately discussing their own restrictions on foreign AI system usage inside China, suggesting the two governments are moving in parallel toward tighter control over which country's AI models companies within their borders can rely on.

For US companies currently running Chinese open-source models in production, the practical question the investigation raises is whether cost and performance advantages will keep outweighing the compliance and reputational risk of using technology now under direct congressional scrutiny. Companies that receive the committees' letters face a choice between defending their technical decisions on the record or beginning a costly migration back to US-origin alternatives.

As first reported by CNBC, the House committees' letters were sent April 29, 2026, with the investigation's findings still pending as of this week.