OpenAI announced several projects with foreign governments this year to facilitate develop so-called “sovereign artificial intelligence” systems. The company says the deals, some of which are coordinated with the U.S. government, are part of a broader push to give national leaders greater control over technology that could reshape their economies.
Over the past few months, sovereign AI has become a buzzword in both Washington and Silicon Valley. Supporters of the concept say it is critical that artificial intelligence systems developed in democracies can spread around the world, especially as China races to deploy its own artificial intelligence technology abroad. “The distribution and dissemination of U.S. technology will prevent our strategic rivals from making our allies dependent on a foreign adversary’s technology,” the Trump administration said in an artificial intelligence roadmap released in July.
At OpenAI, the move also meant working with countries like the United Arab Emirates, which is ruled by a federation of monarchies. OpenAI’s chief strategy officer, Jason Kwon, argues that working with non-democratic governments can facilitate them evolve toward more liberal ones. “The bet is that engagement will be better than containment,” Kwon told WIRED last week at the Curve conference in Berkeley, California. “Sometimes it works, sometimes it doesn’t.”
Kwon’s reasoning is reminiscent of what some politicians said about China more than two decades ago. “We can work to pull China in the right direction, or we can turn our backs and almost certainly push it in the wrong direction” – US President Bill Clinton he said in 2000 when China was preparing to join the World Trade Organization. Since then, many American companies have enriched themselves by trading with China, but the country’s government has only become more authoritarian.
Some argue that true sovereignty can only be achieved when a government is able to control – and to some extent control – a given AI model. “In my opinion, there is no sovereignty without open source,” says Clément Delangue, CEO of Hugging Face, a company that provides open-source artificial intelligence models. China is already ahead in this regard, with its open source models rapidly gaining popularity around the world.
What exactly is “sovereign artificial intelligence”?
Today’s sovereign AI projects range from partial to full control over the entire technology stack, meaning the government manages the entire AI infrastructure, from hardware to software. “The one thing all of these solutions have in common is the legality aspect – by tethering at least some of the infrastructure to geographic boundaries, design, development and implementation will then be compliant with some national regulations,” says Trisha Ray, associate director of the Atlantic Council’s GeoTech Center.
The deal announced by OpenAI in partnership with the US government in the United Arab Emirates includes: 5 gigawatt data center cluster in Abu Dhabi (200 MW of the total planned capacity is to be launched in 2026). The UAE is also rolling out ChatGPT across the country, but it doesn’t look like the government will have the ability to look under the hood or change the chatbot’s inner workings.
Just a few years ago, the idea of building artificial intelligence infrastructure in authoritarian countries could have sparked employee protests in Silicon Valley. In 2019, Google employees pushed away against the tech giant’s plan to implement a censored search engine in China, which would ultimately lead to the project being canceled. “What happens with some LLM projects is quite similar, but doesn’t cause as much opposition,” Ray says. “This belief that ‘well, yes, if you operate within a country’s borders, you have to comply with all the laws of that country’ has become much more normalized over time.”