Three major American security agencies have accused Chinese artificial intelligence companies of copying US-built AI models on a large scale. According to reporting by NOS Tech, the agencies describe these practices as “aggressive and malicious” and say they sit at the core of how Chinese AI technology is being developed.
What happened
NOS Tech reports that three large American security services have leveled a serious charge against several Chinese AI companies. The agencies say that firms including DeepSeek, Moonshot, Alibaba and Z.ai are extracting knowledge on a large scale from powerful American AI models, among them Claude (built by Anthropic), GPT (built by OpenAI) and Gemini (built by Google).
The technical term used in the reporting is “extraction.” According to NOS Tech, this is a method in which a new AI application is built on top of another, more capable AI model. The powerful model generates answers, and those answers are then used to train and develop a new model. The technique is common in AI research and development. What the American agencies object to is the alleged scale and purpose: they say Chinese firms are using it to close the gap between their own systems and the strongest American models.
In the words quoted by NOS Tech, the agencies wrote that “the enormous scale of these operations and their specialised nature indicate that extraction is not a supplement to the development of AI models by these companies, but rather forms the core of it.” That is a strong claim, because it moves the accusation from occasional borrowing to a foundational dependence.
The dispute is not new. NOS Tech notes that American AI companies have complained about copying by Chinese competitors for some time. In February, Anthropic, the company behind Claude, accused the Chinese firms DeepSeek, Moonshot and MiniMax of “illegally extracting Claude’s capabilities to improve their own models.” The matter has also reached the White House. According to NOS Tech, in July a senior adviser to President Trump accused the Chinese company Moonshot of copying AI technology from Anthropic.
China rejects the accusations. NOS Tech, citing the news agency Reuters, reports that a spokesperson for the Chinese government called them false. According to that spokesperson, the quality of Chinese AI technology is the result of “high-level science and technological independence.”
Why it matters in the Netherlands
For internationals living in or moving to the Netherlands, a dispute between Washington and Beijing over AI models can feel distant. It is not. The tools named in the NOS Tech reporting are the same ones many people here use at work and at home every day. Claude, GPT and Gemini power chat assistants, coding tools, translation, and the customer service systems you already encounter. If a legal or political conflict changes how these models are shared, priced or restricted, the effects reach ordinary users, not only the companies building them.
Start with jobs. The Netherlands has a substantial technology sector, with a strong presence of international staff in Amsterdam, Eindhoven, Utrecht and around the university cities. Many software developers, data specialists and product teams here build on top of the American models named in the article. A dispute over how models can be copied or accessed raises questions about which providers a Dutch employer can safely rely on. For an international worker whose role depends on a specific tool, that uncertainty is worth watching.
On costs, the picture is indirect but real. Competition between American and Chinese AI providers has helped push the price of using these models down. Chinese firms such as DeepSeek have drawn attention partly because they offered capable models cheaply. If the accusations reported by NOS Tech lead to restrictions, licensing fights or reduced competition, the price you or your employer pays for AI services could move in the other direction. Cheaper alternatives becoming less available tends to lift the cost of the ones that remain.
Daily life is affected through the products layered on top of these models. Banking apps, health portals, government digital services and the language tools that internationals lean on while learning Dutch increasingly use AI in the background. The Netherlands also operates within the European Union’s AI rules, which set requirements on transparency and risk. A high-profile intellectual property fight of the kind NOS Tech describes feeds directly into the European debate about where AI systems come from and whether their training was lawful. That debate shapes what Dutch and EU regulators demand from the tools you use.
For the Dutch tech sector specifically, the story is a reminder that the country is a consumer of foundation models rather than a builder of the largest ones. Dutch startups and research groups typically build applications on top of American systems. That makes them sensitive to any disruption in access, and it strengthens the argument, already active in European policy circles, for more homegrown or European AI capacity. NOS Tech places this story under its Economy and Tech reporting for good reason: the commercial stakes are as large as the political ones.
What to watch next
The first thing to watch is whether the accusations move from statements to action. So far, according to NOS Tech, we have security agency findings, an earlier complaint from Anthropic, and a White House adviser’s comments. Whether any of this turns into formal legal cases, export controls or trade measures will determine how much ordinary users feel it.
Second, watch the courts. NOS Tech’s related coverage points to a wider wave of AI legal disputes, including music companies Sony and Warner suing Anthropic over copyright, and a German case concerning song lyrics. How judges treat the training and copying of AI models will set precedents that reach into Europe.
Third, watch the European response. If access to leading models becomes more contested, expect louder calls in the Netherlands and the wider EU for regulation and for domestic alternatives. For internationals here, the practical question is simple: will the AI tools you rely on stay available, affordable and legal to use. This story is an early signal that the answer is not guaranteed.
Finally, watch China’s stance. Beijing has firmly denied the claims, so this looks set to remain a live point of friction between the two largest AI powers, with Europe caught in between.
Source: NOS Tech
