Moonshot AI’s debut targets Chinese startups in the US
The United States is intensifying its scrutiny of Chinese artificial intelligence companies, alleging that several are utilising American technology without authorisation and contributing to the enhancement of China’s military capabilities. The latest development comes from Chinese AI startup Moonshot AI, which has recently introduced its Kimi K3 model. This model has garnered global attention due to its impressive performance coupled with a lower cost structure. As Chinese AI models gain traction globally, Washington is contemplating more stringent measures, such as sanctions, export restrictions, and limitations on access to advanced AI chips, according to reports. The Donald Trump administration has issued a warning regarding the potential imposition of sanctions or other restrictions on Chinese AI companies, including Moonshot AI, due to allegations of technology copying from prominent American AI developers. The concerns arose following the launch of Kimi K3 by Moonshot, an open-weight AI model boasting 2.8 trillion parameters. The model has garnered significant attention for delivering superior performance at a more competitive price point compared to numerous rival systems. Moonshot is preparing for a listing on the Hong Kong stock market within the next six months. The company is anticipated to initiate its final funding round prior to the listing in August, with a reported valuation of approximately $50 billion, according to the report.
The report indicated that US officials have asserted that Moonshot may have employed a technique referred to as “distillation” in the construction of its latest model. Distillation is a method of AI training whereby a smaller model assimilates knowledge from the outputs of a larger, more proficient model. The approach is prevalent in the AI industry; however, it raises controversy when companies utilise proprietary models without obtaining permission. Michael Kratsios alleged that Moonshot built an internal system capable of carrying out large-scale distillation of American AI models while making the activity difficult to detect. He also claimed the company had obtained servers powered by the latest GB300 chips and had access to additional computing resources in Thailand for AI training, source reported. Similar allegations have been previously levied by Anthropic and OpenAI against various Chinese AI firms, including Alibaba, DeepSeek, and Minimax. The US government has indicated its readiness to implement more robust measures should it uncover evidence of intellectual property theft. Treasury Secretary Scott Bessent indicated that Chinese firms engaging in the illicit replication of American AI technology may encounter sanctions or be added to the US Entity List. Companies on this list are typically prohibited from utilising advanced American technologies, including Nvidia’s most powerful AI chips. Concurrently, the Bureau of Industry and Security, a division of the US Department of Commerce, has initiated an inquiry into the potential acquisition of restricted US AI chips by Chinese firms via unofficial means. If violations are confirmed, additional Chinese AI firms may encounter stricter export restrictions.
The United States expresses apprehensions as Chinese artificial intelligence firms swiftly broaden their footprint in the international marketplace. Chinese open-source and open-weight AI models have gained traction due to their competitive performance coupled with considerably lower costs. According to a report, OpenRouter, a platform facilitating developer access to numerous AI models via a unified interface, has seen Chinese models represent at least 30 percent of enterprise usage on a weekly basis since February. Large international companies, including Uber and Airbnb, have also recognised their utilisation of certain Chinese open-source AI models. Washington’s concerns extend beyond the realm of intellectual property. US media reports indicate that the White House is considering the imposition of broader restrictions on advanced Chinese AI models due to concerns related to cybersecurity risks. One concern is that AI models could contain hidden “backdoors”—secret access points that might allow malicious actors to exploit systems. The report cited an AI engineer stating that such risks are technically feasible if harmful code or data is integrated during the training of an AI model. However, he noted that many large companies mitigate these risks by operating open-source models on their own isolated servers rather than depending on public internet services or official APIs.
The scrutiny extends beyond AI startups. In June, the US Department of Defence included several prominent Chinese companies, such as Alibaba, BYD, and Baidu, on its roster of firms it suspects have connections to China’s military. The companies are primarily recognised for their consumer-oriented ventures, including e-commerce, electric vehicles, and internet services, rather than for their involvement in defence equipment. However, Washington contends that numerous contemporary technologies can fulfil both civilian and military functions. The increasing emphasis is propelled by the concept of “dual-use technology” — innovations that serve both commercial and defence purposes. Artificial intelligence stands out as a prime illustration. AI has the potential to enhance consumer products and business services; however, its applications extend to military planning, intelligence gathering, cyber operations, satellite image analysis, and autonomous weapons. The same argument holds true for technologies including batteries, robotics, cloud computing, drones, and advanced manufacturing. Products developed for commercial markets may also enhance military capabilities.
US officials contend that these technologies may evolve into strategic assets in forthcoming geopolitical competition. The Pentagon upholds a roster of Chinese enterprises that it contends are associated with or bolster China’s military in accordance with US defence legislation. Inclusion on the list does not inherently prohibit companies from conducting business in the US or enforce comprehensive sanctions. However, the designation may restrict engagement with the US defence sector and heighten regulatory oversight. China has dismissed numerous US designations, contending that Washington is leveraging national security concerns to impose restrictions on Chinese technology firms. Several firms identified by the Pentagon have also refuted any connections to military operations. As artificial intelligence increasingly underpins economic expansion and defence strategies, the technological competition between the United States and China is broadening significantly beyond mere trade considerations.
