Artificial intelligence has become one of the main arenas of technological competition between the United States and China. On September 8, 2026, the U.S. government accused six Chinese artificial intelligence companies, including DeepSeek, Moonshot AI, and Alibaba, of allegedly using a technique known as distillation to obtain capabilities developed by U.S. companies. Washington considers that some of these activities may have reached a scale large enough to represent a systematic effort to extract technology. However, the accusations must still be distinguished from proven facts, since not every use of distillation constitutes theft, and China rejects the U.S. claims.
What Is Model Distillation?
Model distillation is a technique used in the development of artificial intelligence systems that allows a model to be trained using information generated by another model, generally larger or more advanced. In the context of Chinese AI, this technique can be used to develop systems capable of learning certain patterns from already existing models. In simple terms, a system can receive numerous questions and analyze the answers produced by another model in order to learn certain behavioral patterns, something that can also form part of the development of Chinese AI. This can make it possible to develop smaller, specialized, or less costly systems without having to exactly reproduce the entire training process of the original model. The technique is not illegal in itself and can be used legitimately when there is authorization to use the data or generated responses, both in Chinese AI projects and in other artificial intelligence systems.
The problem arises when a company uses another company’s proprietary model in a way that could violate its terms of service or rights. According to the U.S. allegations, some Chinese AI companies may have made large numbers of queries to models developed by U.S. companies in order to collect information that could later be used in their own systems. Washington considers that certain activities related to Chinese AI took place on an industrial scale and that this would not simply involve researchers studying the capabilities of another model. The U.S. concern focuses, in part, on the possibility that Chinese AI companies may use these practices to accelerate the development of their own systems. Nevertheless, establishing that illegal conduct occurred would require knowing precisely how the data was obtained, what access conditions existed, and how it was subsequently used.
The Chinese Companies Mentioned
Among the companies mentioned by U.S. authorities are DeepSeek, Moonshot AI, and Alibaba, three important actors in the Chinese AI sector. Other Chinese companies such as MiniMax, StepFun, and Z.AI, which are part of the growing Chinese AI ecosystem, have also been linked to the allegations. According to the U.S. version, these companies allegedly directed efforts toward systems developed by U.S. companies such as Anthropic, OpenAI, and Google, with the aim of obtaining information about their capabilities. The allegations are particularly relevant because they involve several companies linked to the development of Chinese AI and not just a single isolated company.
The inclusion of DeepSeek is particularly significant because of the international impact the company gained after presenting models capable of competing in certain tasks with systems developed by U.S. laboratories. Its emergence generated an intense debate about the costs necessary to develop advanced artificial intelligence and about the ability of Chinese AI companies to reduce their technological limitations. Moonshot AI has also become a relevant actor through the development of its Kimi models, while Alibaba possesses significant research and development resources in this field. The existence of this ecosystem demonstrates that China has multiple companies capable of participating in the global race for artificial intelligence and strengthening the development of Chinese AI.
The Response from Beijing
China has rejected the accusations made by the United States and has questioned Washington’s interpretation of distillation. From the Chinese perspective, this technique is part of the tools commonly used in the development of artificial intelligence and cannot automatically be considered a form of illegal appropriation of technology, including technology used by Chinese AI companies. Beijing has also accused the United States of using arguments related to national security to limit competition from Chinese technology companies. These statements should be understood within the context of a rivalry that in recent years has expanded from semiconductors to artificial intelligence and other sectors considered strategic for the development of Chinese AI.
The Chinese position is also related to U.S. restrictions on advanced technologies. Washington has imposed export controls on certain chips and equipment used to develop artificial intelligence systems, arguing that some of these technologies may have military applications. China has responded by attempting to increase its domestic capacity to produce semiconductors and develop its own alternatives, elements that are fundamental to strengthening its Chinese AI industry. In this context, the United States considers the possibility that Chinese companies could obtain advanced capabilities from foreign models despite existing restrictions to be particularly concerning. Beijing, on the other hand, maintains that its Chinese AI companies should be able to compete and develop technologies through research methods that are considered legitimate.
The Background with Anthropic
The accusations by the U.S. government also have a background in complaints made by private companies. Anthropic, developer of the Claude model, previously accused several Chinese companies, including DeepSeek, Moonshot AI, and MiniMax, of carrying out large-scale distillation campaigns related to the development of Chinese AI. The company stated that numerous accounts had been used to make large numbers of queries to Claude and obtain responses that could later be used to develop competing systems. Anthropic considered that these activities could allow other companies to significantly reduce the time and resources needed to create certain Chinese AI models.
These allegations are important for understanding the origin of U.S. concerns, but they should not be confused with a judicial ruling. A company may present evidence about activities it considers abusive and request that authorities intervene, but it is up to the competent authorities to determine whether a violation occurred and what its legal nature would be. Nor can it be assumed that all the companies accused carried out exactly the same activities within the Chinese AI sector. The number of queries, the way they were obtained, the subsequent use of the data, and the contractual conditions may vary considerably from one case to another.
A New Form of Technological Competition
The controversy demonstrates that artificial intelligence has created forms of competition different from those that existed in other industries. A competitor can publicly analyze a product, study its results, and use that information to try to improve its own system, something that is also part of the evolution of Chinese AI. In the case of generative models, it is also possible to make enormous numbers of queries and observe how the system responds to different problems. If these queries are carried out automatically and on a large scale, they can provide useful information about the capabilities, limitations, and patterns of a model, an aspect that is relevant to international competition over the development of Chinese AI.
This possibility creates difficulties for developers because a model can reveal part of its behavior every time it responds to a user. Companies that offer commercial models attempt to limit certain uses through security systems and terms of service, both to protect their technologies and from competitors in the Chinese AI sector. They can also detect patterns of queries that indicate automation or systematic extraction of information. However, completely preventing someone from learning from a model’s responses is much more complicated than protecting a file that directly contains code or confidential information, which adds difficulties to the development and regulation of Chinese AI.
The Problem of Intellectual Property
The dispute raises a complex legal question: what exactly does it mean to copy an artificial intelligence model? In other sectors, the appropriation of technology may involve copying blueprints, formulas, source code, industrial designs, or confidential documents. Artificial intelligence models are different because much of their capabilities are represented through enormous numbers of parameters and statistical relationships, a characteristic that is also present in Chinese AI systems. A third party can attempt to reproduce certain behaviors without directly accessing those parameters, using the responses generated by the original system.
For this reason, it is necessary to distinguish between learning from a system and illegally appropriating its technology. Scientific research has always depended on knowledge previously developed by other people and organizations, a dynamic that also influences the advancement of Chinese AI. Published algorithms, scientific papers, and known techniques can be used by researchers around the world within the limits established by the respective licenses and regulations. The situation would be different if a company obtained protected information through deception, unauthorized access, or deliberate violation of the conditions governing the use of a proprietary system.
The Importance of Contracts
Terms of service can become one of the central elements of this conflict. Artificial intelligence providers can establish restrictions on the automation of queries, the mass creation of accounts, and the use of their models for certain purposes, including situations related to the development of Chinese AI. If an organization were to use fake accounts to obtain millions of responses with the intention of training a competing system, there could be relevant contractual issues. However, a possible violation of a contract does not necessarily amount to a crime, and the consequences would depend on the specific circumstances.
This difference is important because the term "theft" can oversimplify a technically complex situation. The current allegations do not necessarily describe a scenario in which a company directly entered another company's servers to copy its models. In certain cases, the issue appears to be related to the mass acquisition of responses through accounts and access to commercial services, a practice that could have consequences for competition between Chinese AI and U.S. systems. Determining the nature of these activities would require technical evidence, access logs, and evidence about how the information obtained was subsequently used.
DeepSeek and China's Advancement
The case of DeepSeek is especially important within this competition because the company demonstrated that advances in artificial intelligence are not concentrated exclusively in Silicon Valley. Its models attracted international attention because of their performance and the strategies used to reduce certain development costs, becoming one of the main references for Chinese AI. The phenomenon raised questions about the ability of the United States to maintain a technological advantage over China indefinitely and about the rapid advancement of Chinese AI. It also fueled a debate over whether chip restrictions can actually prevent Chinese companies from developing competitive systems.
The evolution of DeepSeek does not mean that China has eliminated its technological difficulties. Chinese AI companies still face restrictions related to certain semiconductors and advanced components, in addition to other obstacles resulting from U.S. trade policies. However, the emergence of competitive models has demonstrated that innovation can continue even under conditions of limited access to certain technologies. This makes Chinese AI companies strategic actors for Beijing and explains why their advances are being closely monitored by U.S. authorities.
A Broader Technological Race
The rivalry between the United States and China is not limited to artificial intelligence models or the development of Chinese AI. It also includes semiconductors, data centers, cloud computing services, energy, specialized talent, and access to capital. The country that manages to combine these elements more efficiently could gain a considerable advantage in the development of future generations of artificial intelligence. Therefore, the allegations concerning distillation are part of a much larger competition to control the infrastructure necessary to develop and use advanced systems, a fundamental aspect for both the United States and Chinese AI.
The United States maintains important advantages thanks to its large technology companies, universities, research capabilities, and availability of private capital. China, for its part, has an enormous domestic market, a significant base of researchers, and considerable industrial capacity that favors the growth of Chinese AI. Both countries have incentives to reduce their dependence on each other in sectors considered strategic, including artificial intelligence. The result could be an acceleration of innovation, but also greater fragmentation of the global technology market and increasingly intense competition surrounding Chinese AI.
The Difficulty of Verifying the Allegations
Demonstrating a large-scale distillation campaign requires detailed technical evidence. Investigators may need to analyze query logs, accounts used, traffic patterns, computing infrastructure, and data related to the training of models, including those linked to the development of Chinese AI. Even if it is demonstrated that a company made millions of queries to another system, it would still be necessary to establish what those responses were used for and whether they had any connection to the development of Chinese AI systems. The number of queries may be a relevant factor, but by itself it does not demonstrate that illegal conduct occurred.
It is also necessary to determine whether the Chinese AI companies acted individually or as part of a coordinated strategy. Claims of direct involvement by state agencies require an even higher level of evidence. The existence of Chinese companies benefiting from certain information does not automatically demonstrate that the Chinese government ordered or financed the activities related to Chinese AI. In a geopolitical dispute of these dimensions, maintaining this distinction is essential to avoid turning political allegations into conclusions that have not yet been proven.
What Could Happen from Now On?
The United States could increase pressure on the companies mentioned through investigations, export controls, or other restrictions. China could respond by denouncing these measures as attempts to hinder its technological development and by taking action against U.S. companies, which could also affect the international growth of Chinese AI. There is also the possibility that both governments will hold discussions on security and establish mechanisms aimed at reducing the risks arising from advanced artificial intelligence. The outcome will depend both on the evidence that emerges and on the general evolution of relations between Washington and Beijing, as well as on the position they adopt regarding the development of Chinese AI.
Artificial intelligence companies will also have to adapt. U.S. providers will probably strengthen mechanisms designed to detect automated queries and patterns that may indicate systematic information extraction, even when these activities are related to Chinese AI companies. Companies developing competing models could face greater pressure to explain where their training data comes from and what technologies they use, an issue that is particularly relevant to Chinese AI. Transparency will become increasingly important in a sector where it is difficult to directly observe the entire process through which a model acquires its capabilities.
The U.S. accusations against six Chinese companies represent a new stage in the growing technological rivalry between the United States and China. Washington maintains that certain companies may have used U.S. models through large-scale distillation campaigns, while Beijing rejects this interpretation and defends the use of research techniques that are part of the development of artificial intelligence and Chinese AI. The existence of distillation does not in itself demonstrate a crime, since the technique can be used legitimately under different circumstances. To determine whether there was improper appropriation, it would be necessary to establish how the data was obtained, whether access conditions were violated, and what use was subsequently made of the information.
The conflict also demonstrates that artificial intelligence has become a strategic technology. The United States seeks to preserve the advantage of its companies and limit the transfer of capabilities it considers sensitive, while China is attempting to develop a technological ecosystem capable of competing with the West and reducing its dependence on foreign providers. Companies such as DeepSeek and Moonshot AI are part of this process and contribute to the growth of Chinese AI, although the specific allegations against them should continue to be treated as allegations as long as there is not sufficient public evidence to establish definitive responsibility.
The most important challenge will be finding a balance between technological protection and innovation. Governments have reasons to protect trade secrets and capabilities considered strategic, but excessively broad restrictions could harm research and increase the fragmentation of the industry. It will also be necessary to establish clear rules to determine when learning from a model constitutes a legitimate practice and when it may become improper extraction of technology, an issue that will become increasingly relevant as Chinese AI and artificial intelligence advance worldwide.
For companies seeking to adapt to this technological landscape, having an appropriate strategy for digital transformation, infrastructure, and information technology solutions is becoming increasingly important. ITD Consulting offers specialized IT services to help organizations address the challenges of an ever-evolving technological environment, take advantage of new tools, and strengthen their digital capabilities. If you would like to learn more about its services and discuss how technology solutions can contribute to your company's objectives, you can write to [email protected].
EN 
ES