‘Sovereign AI’ Emerges as a New Battleground in the US-China Technology Conflict

OpenAI has revealed several initiatives this year in collaboration with foreign governments to develop what it terms their “sovereign AI” systems. The company states that these agreements, some of which are being facilitated by the US government, are part of a larger effort to provide national leaders with greater oversight of a technology that could transform their economies.

In recent months, the term sovereign AI has gained significant attention in both Washington and Silicon Valley. Advocates argue that it is essential for AI systems created in democratic nations to expand globally, especially as China accelerates its efforts to export its AI technology. “The dissemination of American technology will prevent our strategic rivals from making our allies reliant on foreign adversary technology,” the Trump administration noted in its AI Action Plan published in July.

At OpenAI, this trend has led to partnerships with countries like the United Arab Emirates, which is governed through a federation of monarchies. OpenAI’s chief strategy officer, Jason Kwon, believes that collaboration with non-Democratic governments can encourage them to adopt more liberal practices. “There’s a belief that engagement is preferable to containment,” Kwon told WIRED during an interview at last week’s Curve conference in Berkeley, California. “Sometimes this approach is successful, and at times it isn’t.”

Kwon’s perspective resonates with sentiments expressed by some politicians regarding China over two decades ago. “We can strive to encourage China toward the right direction, or we can ignore it and likely push it in the wrong direction,” stated US President Bill Clinton in 2000 as China prepared to join the World Trade Organization. Since then, numerous American companies have profited by trading with China, yet the government has only grown more authoritarian.

Critics argue that true sovereignty can only be realized if a government can scrutinize—and to some extent regulate—the AI model in question. “In my view, there is no sovereignty without open source,” emphasizes Clément Delangue, the CEO of Hugging Face, a company that offers open source AI models. In this area, China is already making strides, as its open source models are gaining popularity worldwide.

What Is “Sovereign AI” Really?

Current sovereign AI initiatives vary from providing countries with partial to complete control over the entire technological stack, meaning the government oversees all aspects of the AI infrastructure, from hardware to software. “The common underlying element for all of them is the legal aspect—by tying at least part of the infrastructure to geographical boundaries, the design, development, and deployment would then comply with some national laws,” explains Trisha Ray, an associate director at the Atlantic Council’s GeoTech Center.

The agreement OpenAI announced in collaboration with the US government in the UAE includes a data center cluster in Abu Dhabi with a capacity of 5 gigawatts (200 megawatts of that capacity is expected to be operational in 2026). The UAE is also implementing ChatGPT nationwide, though it seems the government will not have the ability to inspect or modify the chatbot’s functionality.

Just a few years ago, the notion of establishing AI infrastructure in authoritarian regimes might have incited protests among workers in Silicon Valley. In 2019, Google employees protested against the tech giant’s plans to launch a censored search engine in China and succeeded in shutting down the project. “What’s occurring with some of these LLM ventures is somewhat similar, but the backlash isn’t as pronounced,” Ray observes. “The idea that, ‘if you’re operating within a country’s borders, you must adhere to its laws,’ has become increasingly accepted over time.”