US Companies Warn of Chinese Theft of Advanced AI and Security Threat

American artificial intelligence giants such as Anthropic and OpenAI have issued warnings about an alleged systematic theft of their cutting-edge technology by China. The practice, which involves creating cheaper open-source chatbots from sophisticated models, threatens the United States’ leadership in the sector and raises national security concerns. Experts indicate that the Chinese strategy can drastically shorten the technological gap.
Anthropic accuses Alibaba of “blatant theft” in AI technology
In a significant move, San Francisco-based Anthropic sent a letter to the US Congress. The company directly accused Alibaba of “blatant theft” of AI through a technique known as “distillation.” This method uses an advanced artificial intelligence model to train a less capable model without permission, resulting in much lower development costs and time. Anthropic CEO Dario Amodei has been an active voice in these allegations.
OpenAI and Google also flag suspicious activity from China
Months earlier, in February, OpenAI, led by Sam Altman, had already pointed out “continuous parasitism efforts” by the Chinese company DeepSeek in relation to its work. In the same period, Google warned of an increase in “malicious activity” originating not only from China, but also from Iran, North Korea and Russia. These incidents together suggest a coordinated pattern of actions.
How “distillation” undermines US leadership in artificial intelligence
The “distillation” of AI models allows Chinese companies to offer their own chatbots at greatly reduced prices on the global market. This cutthroat competitiveness has attracted U.S. business customers looking to save on the expensive “tokens” needed to operate the most advanced models. Sources close to Anthropic reveal that if China were unable to use this technique, its disadvantage over US frontier models would go from 6-9 months to over 18 months, a crucial difference in the pace of innovation.
Rise of Chinese models in the global market
The effectiveness of the Chinese strategy is visible in the global AI landscape. Currently, six of the ten most used artificial intelligence models were developed by Chinese technology companies. These companies are quickly achieving advanced capabilities, comparable to renowned models like Anthropic’s Claude or OpenAI’s ChatGPT.
- Z.AI
- Minimax
- DeepSeek
- Other Chinese companies with rapid development
Implications for national security and the American response
Concerns about AI theft are not limited to the commercial aspect. Powerful Anthropic models, such as “Mythos” and “Fable”, were even temporarily disabled due to White House concerns about their possible use as a weapon by China. National security officials, including from the Trump administration, have been regularly briefed on distillation attacks. One of the sources indicated that “basically all parts of the government,” including the White House and Congress, have been communicated with on the issue. If attacks are not contained, there is a real risk that US global leadership in AI development will be eroded.
Next steps and the future of AI competition
The continuation of these cyberattacks and the advancement of Chinese models could further facilitate the spread of terrifying hacking skills and the development of autonomous weapons. Anthropic’s letter to Congress could be a catalyst for a more formal and rigorous response from the US government. The outcome of this technological dispute will have a profound impact on geopolitics and the world economy in the coming years.






