DW Asks: Is China's Rapidly Advancing AI 'Cheating'?

Chinese artificial intelligence platforms are rapidly narrowing the gap with their American counterparts in cutting-edge capabilities, triggering fresh alarm in the Trump administration and intensifying accusations of intellectual property theft. Is China 'cheating' to catch up? And is the outlook for the US to maintain its lead optimistic?

It seems almost overnight that China has gone from trailing the US by as much as five years in top-tier artificial intelligence (AI) to lagging by just a few months.

US companies like Anthropic's Claude and OpenAI's ChatGPT have poured hundreds of billions of dollars into pushing the frontier of AI capabilities, while China has taken a radically different path.

Faced with US export restrictions on high-end AI chips—critical for the most advanced AI reasoning tasks—Chinese tech firms have shifted focus to developing open-source AI models based on 'good enough' technology.

Platforms including DeepSeek, Alibaba's Qwen (Tongyi Qianwen), and Xiaomi's MiMo are now widely used across Chinese enterprises, government services, and consumer electronics.

China's AI models closing the gap with the US

Chinese startup Moonshot AI released its Kimi K3 model last week, matching the performance of Anthropic's state-of-the-art Claude Fable 5 in multiple key benchmark tests.

In some tests, Kimi K3 not only matched but even surpassed Claude's performance, while requiring less computational power and costing only a fraction of the latter.

Global tech expert Alvin W. Graylin believes Kimi K3 has transformed high-end AI from an expensive luxury into a practical, affordable tool.

Graylin told Deutsche Welle that Kimi K3 has achieved 'commoditization of frontier models.'

He said: 'For 98% of users, that 1% to 2% performance difference isn't worth paying 10 to 20 times more.'

Washington expresses concern over AI 'distillation' technique

China's latest AI advances have triggered a strong reaction in Washington. President Trump is now threatening to ban the use of Chinese AI models in the US.

Meanwhile, Treasury Secretary Scott Bessent warned on Tuesday (July 21) that the US might impose sanctions on Chinese AI models over suspected intellectual property violations.

Anthropic and OpenAI have accused Chinese research institutions of using their models to train their own AI systems—a process known as 'distillation.'

While AI labs around the world use 'distillation' to some extent, these companies say their models are being bombarded with massive query requests to extract advanced reasoning capabilities. The White House warned in April that China is doing this at an 'industrial scale.'

Beijing has pushed back, dismissing the allegations as baseless and calling US threats of sanctions and model bans a form of 'technological bullying' aimed at curbing China's rise.

Julia Neaher, a research analyst at the Stimson Center's Strategic Foresight Hub, believes distillation played a limited role in Kimi's progress.

She told Deutsche Welle: '(Kimi's success) is largely due to talent, infrastructure, and the massive resources invested in model development.'

Why China's open-source model is becoming increasingly popular

In the early days of modern AI, the US chose to keep the inner workings of its leading models closed.

This approach allowed tech giants to maximize profits through subscription services, enterprise licensing, and per-token pricing—the token being the basic unit of text processed by the model.

In contrast, many leading Chinese AI developers have opted for open-source models. This means businesses and developers can download the models, run them on their own servers, and fine-tune them to their needs.

Graylin told Deutsche Welle: 'Open source allows Chinese labs to leverage the global AI community to jointly improve and test models.' He also noted that US chip restrictions have forced Chinese companies to become 'more creative and efficient.'

Washington continues to bet on frontier AI

Despite China's advantages in pricing and deployment, Washington still sees maintaining leadership in frontier AI as critical—because whoever controls the most advanced AI systems could dominate military, economic, and technological competition for years to come.

Frontier AI models can also drive breakthroughs in scientific research and medicine.

While distillation may have helped some Chinese AI models enter the global top 10, Neaher compares the approach to 'teaching students to pass exams, not to truly understand knowledge.'

She told Deutsche Welle: 'Your exam scores might be excellent, but that doesn't necessarily reflect how much knowledge you've actually mastered.'

Neaher's research focuses on AI policy and governance. She expects US hostility toward China's AI development to escalate as Washington re-evaluates its previously 'laissez-faire' approach to AI policy.

Can China overtake in the AI investment boom?

The AI race could also spark a new competition—over which successful AI companies go public to raise capital.

Some US AI companies are already approaching trillion-dollar valuations. Reports suggest Anthropic and OpenAI both plan to launch initial public offerings (IPOs) later this year to capitalize on massive market enthusiasm for AI.

On the Chinese side, DeepSeek and Moonshot AI are also seeking listings. Moonshot plans to go public in Hong Kong by early next year.

Given that Chinese tech firms appear to remain undervalued by the market, US institutional investors have quietly increased their allocations to Chinese tech stocks—seeking to avoid risks from a potential US AI bubble while capturing gains from China's rapid AI rise. In this race, the US and China are rapidly forming a 'neck-and-neck' situation.

Graylin believes: 'The long-term trend is not optimistic for the US.'

He said: 'The US AI market has a serious bubble, and a correction is likely within the next three to twelve months.'

'In contrast, Chinese companies are developing models comparable to US frontier models at one-tenth the cost, and their product prices are also one-tenth of US levels.'

DW Chinese has an Instagram! Search dw.chinese for more in-depth, easy-to-understand graphics and video reports.

© 2026 Deutsche Welle Copyright: All content in this article is protected by copyright law. Unauthorized use without special authorization from Deutsche Welle is strictly prohibited. Any improper conduct will result in compensation and criminal prosecution.