Washington's acute AI paradox: blame China but distill its models: China Daily editorial

A growing narrative in the political circle of the United States suggests that Beijing is behind the mounting local opposition to AI data centers nationwide. The evidence? None, beyond the unsubstantiated notion that communities "giving us trouble" must somehow be acting at China's behest.

The episode exposes an unsettling feature of today's US political discourse: allegations against China need not be substantiated to shape public perceptions. Even though the perceptions might be proved wrong. Fox News recently issued a rare on-air apology after one of its regular guests, Canadian investor Kevin O'Leary, alleged that Chinese money was funding anti-data-center groups in Utah. O'Leary later admitted he had no evidence. Local organizations rejected the accusations outright.

The release of China's Kimi K3 model reinforced a growing realization that the US' dominance in AI is no longer assured. Casting China as a malign actor is politically easier than confronting an increasingly competitive technological rival.

Yet while some in Washington try to curb China's AI industry, they are quietly acknowledging its prospects.

On Tuesday, senior US officials met executives from multiple US tech giants, including OpenAI, Anthropic and Google, to finalize a new voluntary AI governance framework. The most striking feature was what it permits: Chinese open-weight AI models will not be subject to government safety testing under the new regime. In practice, US companies are being given considerable freedom to use, fine-tune and distill Chinese open-source models. That is an implicit admission that Chinese open-weight AI has become too important to ignore.

The contradiction is difficult to miss. Washington accuses China of threatening US AI infrastructure while simultaneously allowing US companies to draw on Chinese AI innovations to enhance their own products. Chinese technology is portrayed as dangerous in political speeches but valuable in commercial practice.

For years, the US enjoyed a lead in most AI-related sectors, from advanced semiconductors and frontier models to cloud infrastructure and developer ecosystems. That lead is narrowing. Chinese companies such as DeepSeek and Moonshot AI have advanced rapidly through iteration, open-source collaboration and large-scale industrial deployment, rather than relying solely on proprietary models.

Washington's response has been a three-pronged containment strategy: restricting China's access to advanced AI chips, limiting semiconductor manufacturing capabilities and tightening controls on cloud-based computing services. These measures are proving less effective against open-source innovation, where knowledge spreads far more rapidly than hardware.

While chips and manufacturing equipment can be restricted, algorithms shared openly across global developer communities are far more difficult to contain. US companies certainly know that. Many reportedly opposed tighter restrictions on open-weight models, warning that excessive regulation would simply hand leadership in the global open-source ecosystem to China.

The White House appears to have accepted that argument now after vowing to ban China's open-source AI. Yet in doing so, it has opened another front in the competition. US technology giants — with unmatched cloud infrastructure, computing resources and developer communities — can now absorb Chinese open-source innovations, integrate them into proprietary platforms and commercialize them at an enormous scale. Openness creates influence, but it also creates opportunities for competitors.

China needs to move beyond releasing powerful models toward building a complete innovation ecosystem encompassing chips, software, data, cloud services and industrial applications.

But China's AI sector still faces some structural constraints. Efficient Chinese-native tokenization remains underdeveloped, AI chip development cycles remain long and capital-intensive and supplies of high-quality training data are expected to tighten in the coming years. Meanwhile, the US retains advantages in cloud computing, venture capital, software ecosystems and its capacity to commercialize emerging technologies. Over the long run, ecosystem builders — not hardware suppliers alone — are likely to capture the greatest value.

Hardware breakthroughs matter, but they are only one element of lasting competitiveness. The real prize is a self-reinforcing AI ecosystem in which chips, models, cloud infrastructure, applications and talent continuously strengthen one another. US restrictions are a serious challenge. But the greater danger is mistaking temporary success of one or two models for lasting competitiveness. In AI, the decisive contest will not be won by the country that delivers the next breakthrough, but by the one that builds the most resilient ecosystem.