The clash between China and the United States in the field of artificial intelligence just got a brand new and explosive chapter.
Three of the most powerful American federal agencies — the FBI (Federal Bureau of Investigation), the NSA (National Security Agency), and CISA (Cybersecurity and Infrastructure Security Agency) — have leveled serious accusations against Chinese AI companies, alleging they have been engaging in what the agencies called industrial-scale theft of American models through a technique known as distillation.
On the other side, China didn’t stay quiet. 😤
The Chinese government fired back at full force, calling the allegations baseless and warning of possible countermeasures if the U.S. pushes forward with restrictions.
And all of this is unfolding at a politically delicate moment: President Xi Jinping is about to visit the United States for a summit with Donald Trump, with artificial intelligence squarely on the agenda.
What looks like a technical fight over how AI models learn is actually much bigger than that. This is a battle for control of the global technological future. 🌐
What is distillation and why did it become the center of this controversy?
To understand the full scope of this mess, you first need to understand what model distillation actually is. This technique, widely known in the artificial intelligence community, basically involves using the outputs of a larger, more powerful model — called the teacher model — to train a smaller, more efficient model, known as the student model.
The result is an AI system that learns to mimic the behavior of the original model without necessarily having direct access to its weights or source code. It is a legitimate approach, extensively published in scientific papers and used by companies all over the world, including right here in the United States.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
There is an important technical detail worth mentioning here: AI models work with what we call tokens, which are the smallest units of text that a system reads or generates. In the joint statement released by the American agencies, the accusation is that Chinese companies like DeepSeek, Alibaba, and Moonshot AI allegedly extracted billions of tokens from American models since at least late 2024, possibly with the knowledge of the Chinese government itself.
The problem, according to the American agencies, lies in how this technique is allegedly being applied. The central claim from the FBI, the NSA, and CISA is that certain organizations linked to China have been using distillation in a systematic and intentional way to extract the knowledge embedded in cutting-edge American models — without authorization, without licensing, and without any form of compensation or transparency.
In practice, it would be like using an American AI as a private tutor to build a competing Chinese AI, for free and in secret, bypassing all the legal and commercial barriers that would normally exist in this process. According to the American statement itself, the scale and sophistication of these campaigns suggest that distillation was not just a supplement to the development of these Chinese systems, but rather the critical core of the entire operation.
China’s response and Beijing’s direct message
When the American agencies made these accusations public, the Chinese government didn’t take long to respond. The Chinese Ministry of Commerce characterized the allegations as part of an American attempt to establish a monopoly over the artificial intelligence industry, making it clear that China will not accept pressure passively.
In a statement sent to the Associated Press, the ministry said that if the United States suppresses Chinese AI companies using distillation as a pretext, China will take firm and decisive countermeasures. The ministry also reinforced a central argument in China’s defense: distillation is a common practice across the entire artificial intelligence industry, including among American companies themselves.
The Chinese Ministry of Foreign Affairs echoed the same message. Spokesperson Mao Ning urged the United States to avoid making baseless accusations or defaming the country. According to her, the development of Chinese AI is the result of high-level technological self-reliance and independent effort, not copying. The representative also argued that all parties should strengthen cooperation to promote AI that is open, inclusive, universally beneficial, and oriented toward the common good.
Xi Jinping’s visit and what’s at stake diplomatically
This tense scenario is playing out at a particularly sensitive time. The American statement came just weeks before President Xi Jinping’s visit to the United States, where Donald Trump is expected to host a summit with a wide range of topics on the table — and artificial intelligence is among the top ones.
In a rare move, Xi Jinping may arrive in the U.S. accompanied by a large number of Chinese executives, according to Reuters. It is still unclear which names would be part of this delegation, but previous meetings between the two countries suggest that AI industry leaders could be in the group. It is worth remembering that in 2015, Xi brought Jack Ma, the founder of Alibaba, with him during a visit to the United States.
On the American side, when Trump met with Xi during a trip to Beijing in May, the U.S. president was accompanied by more than a dozen American executives, including Elon Musk, CEO of Tesla. In other words, the presence of tech giants at these high-level meetings has practically become a tradition.
It is no exaggeration to say that what happens in Washington’s meeting rooms could have a direct impact on the pace of AI development in the years ahead — for American companies and for startups and researchers around the entire planet.
Washington’s hard-line stance
Under Trump’s leadership, the American government has been taking a very firm position against efforts to advance Chinese AI models. And for good reason: China is investing heavily in artificial intelligence and robotics, pouring nearly 300 billion dollars into what the government sees as the country’s next economic engine.
This competitive drive was made clear during a speech by U.S. Treasury Secretary Scott Bessent at an event in Dallas. He stated bluntly that China will never surpass the United States in AI. According to Bessent, the Chinese distill American models precisely because they can never get ahead on their own.
So far, the Chinese companies named in the American statement have not publicly responded to requests for comment.
The real impact on the artificial intelligence industry
Regardless of how this diplomatic dispute plays out, the impact on the artificial intelligence sector is already being felt. American companies are reviewing their API access controls, building mechanisms to detect suspicious usage patterns, and working with regulators to create more specific legal frameworks to protect AI models against so-called model theft.
This could make access to AI tools more restricted, more expensive, or more bureaucratic for independent developers and researchers around the world — a side effect that concerns a significant portion of the technical community.
On the other hand, some argue that this debate is accelerating a conversation that needed to happen. Model distillation is a powerful technique that democratized access to AI, allowing organizations with fewer computational resources to develop competitive systems. If it starts being used as an instrument of technological espionage on an industrial scale, the market needs clear rules — and fast.
Without a minimally agreed-upon international regulatory framework, this technological race between China and the United States could end up hurting the smaller players who benefited the most from the openness of recent years.
There is also a less discussed but equally important dimension: what this episode reveals about the structural vulnerability of AI models as intellectual property. Unlike traditional software, where the code can be protected through encryption and technical licensing, a language model exposed via API constantly delivers fragments of the knowledge that was incorporated during training.
Each generated response is, in a way, a window into the intelligence embedded in that system — and protecting that intelligence against systematic extraction is a challenge the industry has yet to solve in a satisfactory way.
What is at stake here goes far beyond a fight between tech companies or a passing diplomatic crisis. The way China and the United States decide to handle the issue of distillation and model theft will shape the rules of the game for the entire artificial intelligence industry for decades to come.
And everyone — from major corporations to independent developers, from governments to everyday users — will feel the effects of these decisions. This is definitely one to keep an eye on over the coming months. 🤖
