US lawmakers have announced a joint investigation into the security risks posed by Chinese-developed artificial intelligence models. The probe, led by House Select Committee on China Chairman John Moolenaar and House Committee on Homeland Security Chairman Andrew R. Garbarino, will examine the national security and cybersecurity implications of American companies adopting AI systems created by Chinese firms such as DeepSeek, Alibaba, Moonshot AI and MiniMax.

According to the Select Committee on the CCP, these models are low-cost, open-weight and API-accessible, making them attractive to US businesses but potentially dangerous due to their origin.

The committees highlighted that companies like Airbnb and Anysphere have built products using Chinese AI models, raising concerns about hidden vulnerabilities and data exposure. They warned that these systems are trained under China’s censorship regime and could be compelled to share data with the Chinese government under Chinese law.

Chairman Moolenaar stressed that such reliance on Chinese AI threatens critical infrastructure and risks placing sensitive American data in adversary hands. He emphasised that the committees will question Airbnb and Anysphere to understand their decisions and ensure protection against CCP-linked threats.

Chairman Garbarino underscored that the Chinese Communist Party is attempting to leverage America’s AI breakthroughs for Beijing’s strategic advantage. He pointed to adversarial model distillation and the rapid global spread of PRC-developed models as methods to weaken US leadership, undermine trusted American alternatives, and embed CCP-aligned technology across vital software supply chains.

He cautioned that American firms cannot afford to treat Chinese AI as a convenient option when the consequences may include compromised systems, exposed data and long-term dependence on adversary-controlled technology.

The investigation comes amid growing concerns that Chinese AI companies may be using unauthorised model distillation and other unlawful techniques to extract capabilities from advanced American frontier models.

These capabilities are then repackaged into cheaper systems that lack the safety protections of the original US models. Such systems are subsequently marketed to American businesses, developers and consumers, raising alarms about intellectual property theft, cybersecurity risks and supply-chain vulnerabilities.

While model distillation can be a legitimate AI development method, the committees noted that when it is carried out through fraudulent accounts, proxy networks, or violations of US companies’ terms of service, it poses serious threats to national security and innovation.

ANI