Chinese AI companies are accused by U.S. firms like Anthropic and OpenAI of copying proprietary AI models through “distillation,” a technique that enables smaller, cost-effective models to replicate much of the performance of larger ones, fueling rapid growth and billion-dollar valuations in China’s AI sector. This shift challenges the traditional economics of AI development, raising intellectual property concerns while prompting questions about the future value and competitiveness of closed-source versus open-source AI models globally.
Chinese AI companies are facing accusations from American firms like Anthropic and OpenAI of copying proprietary AI models through a technique called “distillation,” which involves training smaller models on outputs from larger, proprietary ones. Anthropic alleges that companies such as DeepSeek, MiniMax, and Moonshot AI used millions of fraudulent accounts to extract capabilities from its Claude chatbot. OpenAI has also raised concerns with U.S. lawmakers about similar practices. Meanwhile, Google’s Threat Intelligence Group has warned about increasing distillation attacks targeting its Gemini model. Despite these allegations, the accused Chinese firms have not publicly responded.
The competitive landscape in AI is shifting as several Chinese AI models have reached capabilities comparable to American counterparts. Many of these Chinese models are open source and more cost-effective, challenging the expensive economics of AI development in the U.S. Experts note that distillation allows smaller models to achieve 80-90% of the performance of larger models with significantly less computing power, enabling faster and cheaper AI solutions. This has led to increased investor enthusiasm in Chinese AI companies, with firms like MiniMax and Z dot AI going public and creating new billionaires.
The wealth generated by Chinese AI founders is notable, with figures such as MiniMax CEO Wang Junjie and Z dot AI chairman Lu amassing fortunes comparable to those of Anthropic’s billionaire co-founders. Other Chinese AI entrepreneurs, including Tsinghua University professor Tong Ji and newcomers like Liang Wangfeng and Yang Zhilin, are also joining the ranks of billionaires. This surge in wealth reflects the rapid growth and investor confidence in China’s AI sector, even as American companies grapple with intellectual property concerns.
From an economic perspective, distillation is a game-changer because it drastically reduces operating costs and accelerates price competition. For example, Z dot AI’s GLM-5 model reportedly costs significantly less per token than Anthropic’s Claude Opus 4.6. While distillation requires a strong base model and skilled engineering, it lowers barriers to entry and flattens competition. Many top-performing open-source models now come from China, and a majority of American startups using open models rely on Chinese technology, facilitated by easy switching through cloud platforms.
Looking ahead, the AI industry faces uncertainty about where long-term value will reside in a world dominated by open models. Closed-source companies like OpenAI and Anthropic, along with startups focused on safety and advanced AI, must demonstrate clear economic advantages to justify their valuations. Meanwhile, stock prices of Chinese AI firms have shown volatility, with some trading at extremely high multiples before pulling back. The evolving dynamics highlight the challenges and opportunities in a rapidly changing global AI market.
