Divided Silicon Valley slows White House plans to ban Chinese AI

The US government has temporarily shelved plans to ban Chinese open-source AI models, after part of Silicon Valley pushed back against the measures. The New York Times reported the development. The potential measures under consideration included sanctions and prohibitions on offering Chinese AI through American cloud platforms.

The tech industry proved internally divided. OpenAI and Anthropic advocated for restrictions; Nvidia, Google and Meta were opposed. After those companies applied pressure on Washington, the administration postponed the plans. A decision is expected before the state visit of Chinese President Xi Jinping, which is scheduled for September.

What was on the table

According to the New York Times, the Trump administration discussed two types of measures. The first was imposing sanctions on Chinese providers of open-source AI models. The second was a cloud ban, under which American cloud providers such as Amazon Web Services, Microsoft Azure and Google Cloud would no longer be permitted to offer Chinese AI models to their customers.

Chinese open-source models, including those from DeepSeek, are now widely available through Western cloud platforms and are used by companies around the world. A cloud ban would block that access and force providers to remove or modify existing services.

The discussion is taking place against the backdrop of a broader American debate about technological decoupling from China. The US government has previously imposed export restrictions on advanced chips, but open-source software falls outside that existing regulatory framework.

Who was for and who was against

On the side of those in favour of restrictions were OpenAI and Anthropic, two of the most prominent American providers of closed AI models. Both companies compete directly with Chinese alternatives and have a commercial interest in making those alternatives less accessible to American customers.

Ranged against them were Nvidia, Google and Meta. Nvidia sells chips that are widely used for training and running AI models, including open-source variants. A ban on Chinese models could dampen demand for those applications. Google and Meta themselves also offer open-source models and have an interest in an open ecosystem. Meta publishes its Llama models as open-source and positions itself as a counterweight to closed models.

The divide reflects a broader philosophical and commercial rift within the AI sector, between companies that view open-source as a risk and those that see it as an opportunity.

Timeline and political context

The New York Times reported that the companies opposed to a ban applied enough pressure to persuade the administration not to proceed with the plans for the time being. A final decision is scheduled for September, when Chinese President Xi Jinping is expected to visit the United States.

That timeline is politically significant. Measures announced just before or during such a visit carry a different diplomatic weight than quiet regulatory decisions. The timing suggests that the administration also views the issue as a negotiating instrument within the broader trade relationship with China.

The debate over Chinese AI models does not stand in isolation either. In recent months, Washington has tightened export rules for AI chips on several occasions. The question of whether software, and specifically open-source models, should be subject to comparable export control mechanisms represents the next step in that policy trajectory.

What this means for the Dutch and European AI scene

For European companies and policymakers, the outcome of this American debate is directly relevant. European cloud providers and AI startups make use of American infrastructure, which means they could potentially fall within the scope of a future cloud ban. If American platforms are required to remove Chinese models, that will have consequences for European customers operating through those same platforms.

At the same time, the American discussion is placing pressure on European AI policy. The EU is developing its own regulatory framework through the AI Act, but has not yet taken a position on specific restrictions targeting Chinese open-source models. However Washington ultimately decides, that decision could influence the political conversation in Brussels and The Hague about what is permissible in public and private AI applications. For Dutch and European founders using Chinese open-source models in their products, it is advisable to monitor policy developments in Washington closely over the coming months.