What Happened
On April 29, 2026, House Homeland Security Committee Chair Mark Green and Select Committee on China Chair John Moolenaar jointly signed two investigation letters, delivered to Airbnb and Anysphere (Cursor’s parent company), demanding detailed responses within 30 days regarding their use of Chinese AI models.
The core allegations in the investigation letters are as follows:
| Subject | Chinese Model Used | Investigation Focus |
|---|---|---|
| Airbnb | Alibaba Qwen | Data security, model distillation, user privacy |
| Anysphere (Cursor) | Moonshot AI Kimi | IP leakage risk, code supply chain security |
The investigation letters explicitly state that both companies’ decisions may expose American enterprise and user data to “unacceptable national security risks.” The committee requires companies to disclose:
- Specific reasons and decision-making processes for choosing Chinese models
- Whether data is transmitted to servers within mainland China
- Whether Chinese models are used for model distillation or fine-tuning
- Internal security assessment reports and risk mitigation measures
Why This Is a Watershed Moment
This is not the first time the US has expressed concerns about AI security — but it is the first formal investigation by legislators targeting specific companies’ use of Chinese AI models.
Previous AI decoupling measures focused on:
- Chip export controls (Nvidia A100/H100 sales ban)
- Investment restrictions (executive orders banning investment in Chinese AI companies)
- Government agency bans on Chinese AI applications
The nature of this investigation is fundamentally different: it targets private companies’ technology selection decisions, focusing on software-level model usage rather than hardware-level chip procurement.
This means the US is building a new regulatory framework: not only restricting the inflow of Chinese AI hardware, but also scrutinizing the use of Chinese AI software.
The Situation for Both Companies
Airbnb
Airbnb integrated Alibaba’s Qwen API into its internal systems, primarily for multilingual customer service and content moderation. The investigation committee believes these scenarios involve large amounts of personal user information, creating cross-border data transmission risks.
Notably, Qwen’s global API services are provided through Alibaba Cloud’s international nodes, and data theoretically does not flow back to mainland China. But the investigation letter questions precisely this “theory” — whether Chinese enterprises could potentially access user data during model training or inference.
Anysphere (Cursor)
Cursor is the world’s most popular AI-powered coding IDE, with millions of developer users. Anysphere was exposed to be using Moonshot AI’s Kimi model in its backend model routing, handling code completion requests for some users.
Cursor’s business model is built on “helping developers write code more efficiently,” yet it uses Chinese models at the infrastructure level — this structural contradiction is exactly what the investigation targets.
Impact on Chinese AI Companies
Short-term impact is limited, but the signal is significant:
- Rising compliance costs for overseas expansion: Any Chinese AI company wanting to enter the US market must prepare in advance for similar scrutiny
- Model distillation risk is now quantified: US companies using Chinese models for distillation is now explicitly defined as a security threat
- The “cheap but safe” narrative is challenged: Chinese models have won on cost-performance, but security reviews may now become a hidden cost
Action Recommendations
For developers using or considering integrating Chinese AI models:
- Enterprise users: Immediately audit existing AI supply chains, assess whether Chinese models are involved, and prepare compliance documentation
- Individual developers: Monitor legislative developments and consider using locally deployed or open-source models as alternatives
- Chinese AI companies: Accelerate independent operations of overseas data centers, achieving complete isolation of data storage and processing
Landscape Assessment
This investigation is just the beginning. As AI models become deeply embedded in core business operations, model supply chain security will become as important a policy issue as chip export controls.
For Chinese AI companies, the “cheap and good” product advantage is being offset by geopolitical factors. The key to future competition is not just model capability, but trust-building capability.