'America First policy unintentionally pushing US allies toward sovereign AI,' warns Coursera co-founder

Andrew Ng, co-founder of Coursera and one of the most influential voices in artificial intelligence, has warned that US policies are unintentionally pushing allies away from American AI technology — accelerating global interest in what is increasingly called sovereign AI. 

In a detailed post on X (formally twitter), Ng argued that years of policy decisions across multiple US administrations have made other nations wary of over-reliance on American technology, especially in a domain as strategically critical as AI. 

“Sovereign AI,” Ng explained, refers to a nation’s ability to access and deploy artificial intelligence without depending on foreign powers that could restrict or cut off access. While the concept remains loosely defined, its appeal is growing rapidly as governments seek technological autonomy. 

Policy decisions and eroding trust 

Ng pointed to several US actions that, in his view, undermined global confidence. Following Russia’s invasion of Ukraine, US sanctions led to ordinary Russian consumers losing access to credit cards — demonstrating how geopolitical decisions can directly affect civilian technology access. 

More recently, export controls introduced under President Joe Biden limited AI chip sales to many countries, including US allies. Under President Donald Trump, Ng said, the “America First” approach further intensified concerns through broad tariffs, aggressive rhetoric toward allies, and a hostile immigration environment. 

He also cited globally circulated footage of U.S. immigration enforcement actions, claiming they have deterred skilled professionals from traveling to the country out of fear of arbitrary detention. 

Why Sovereign AI is gaining momentum 

Given AI’s importance to economic competitiveness and national security, Ng said governments increasingly want guarantees that no single foreign actor can restrict their access to critical models or infrastructure. 

Complete independence, however, is unrealistic. Advanced AI chips are largely designed in the U.S. and manufactured in Taiwan, while much of the hardware supply chain runs through China. Instead, countries are seeking alternatives to frontier AI models from US firms such as OpenAI, Google, and Anthropic. 

This shift has helped drive adoption of open-weight models developed in China — including DeepSeek, Qwen, Kimi, and GLM — particularly outside the United States. 

Open source as the strategic shortcut 

Ng emphasized that sovereign AI does not require nations to build everything themselves. Participation in the global open-source ecosystem, he said, offers the most cost-effective path to autonomy. 

Just as countries rely on Linux, Python, and PyTorch — software no single nation controls — open-weight AI models allow governments to ensure uninterrupted access while remaining at the cutting edge. 

This has already spurred action. Ng noted recent initiatives such as the UAE’s launch of K2 Think, an open-source reasoning model, and highlighted efforts underway in India, France, South Korea, Switzerland, Saudi Arabia, and others to build domestic foundation models and secure trusted compute infrastructure. 

Fragmentation, competition, and an irony 

While Ng warned that global fragmentation and declining trust among democracies is harmful, he also pointed to a potential upside: increased competition. 

He compared the situation to web search, where US giants like Google and Bing dominate globally, but regional players such as Baidu and Yandex succeeded locally. A similar pattern in AI, he suggested, could slow consolidation, foster innovation, and reduce dependence on a small group of dominant firms. 

Ng concluded with an irony: policies intended to prioritize US interests may ultimately expand global access to AI by encouraging open source development and strengthening non-US ecosystems.