US Lawmakers Investigate Chinese AI Firms for Security Risks
House Select Committee on China Chairman John Moolenaar and House Committee on Homeland Security Chairman Andrew R. Garbarino have initiated a joint investigation into the national security and cybersecurity risks associated with the increasing deployment of Chinese-developed artificial intelligence models. These encompass low-cost, open-weight, and API-accessible systems developed by Chinese companies including DeepSeek, Alibaba, Moonshot AI, and MiniMax, as stated. The choices made by Airbnb and Anysphere to develop their offerings utilizing Chinese Communist AI frameworks pose a risk to essential infrastructure relied upon by Americans on a daily basis. The artificial intelligence models employed by these companies are shaped by the censorship practices in China, thereby embedding concealed weaknesses that jeopardize the data and operations of American enterprises.
Chinese AI companies operate under the jurisdiction of Chinese law, which may compel them to provide data collected from platforms like Airbnb and Anysphere to the Chinese government upon request. Chairman Garbarino and I anticipate insights from Airbnb and Anysphere as we delve into their actions, and we will persist in our collaborative efforts to safeguard the American populace from CCP threats,” stated Chairman Moolenaar. The Chinese Communist Party is attempting to leverage America’s advancements in AI to gain a strategic edge for Beijing. Through adversarial model distillation and the swift global dissemination of PRC-developed models, China is strategically positioning itself to undermine US leadership, diminish reliable American alternatives, and integrate CCP-aligned technology throughout the software supply chains that are crucial to our economy and national security.
American companies must recognize that treating Chinese AI merely as a cost-effective and convenient resource carries significant risks, including the potential for exposed data, compromised systems, and a lasting reliance on technology controlled by adversaries. Chairman Garbarino expressed pride in collaborating with Chairman Moolenaar on this investigation aimed at gaining insights into the adoption of these models in the United States and identifying necessary measures to safeguard American innovation and national security, as stated in SCCCP’s release. The investigation arises from growing apprehensions that AI firms operating in China could be employing unauthorized model distillation and other illicit techniques to extract functionalities from sophisticated American frontier models, subsequently reconstituting those functionalities into more affordable systems that lack the safety measures inherent in the original US models.
According to the SCCCP release, these are subsequently marketed or made accessible to American businesses, developers, and consumers. Model distillation can indeed serve as a valid approach in the realm of AI development. However, the SCCCP release highlighted that when distillation is executed via fraudulent accounts, proxy networks, circumventing access controls, or breaching the terms of service of US companies, it raises substantial concerns regarding model origin, intellectual property, cybersecurity, and supply-chain security.
