South Korea's Sovereign AI Gambit: A High-Stakes Experiment in Autonomy

The Diplomat

a day ago

South Korea's new Lee Jae-myung administration made establishing the country as a global AI leader one of its top priorities and took immediate action upon assuming office in June. Lee appointed Ha Jung-woo, the former AI chief of Korean tech giant NAVER and the nation's most vocal advocate for the so-called 'sovereign AI' discourse, as South Korea's first-ever presidential chief AI adviser. Days later, Bae Kyung-hoon, who led another massive AI foundation model development project at LG, was nominated as the new minister of science and ICT.
The appointments were unconventional in several respects. Instead of the politicians or senior academics who typically rise to high administrative posts, two 40-something industry experts – considered young blood in South Korea's political scene – were brought to the forefront. More importantly, they were the very executives who had led South Korea's two largest large language model (LLM) projects. Their appointments are backed by a 100 trillion won (approximately $75 billion) national investment plan that Lee had pledged since his campaign. This was a clear message: Seoul has bet its future on building sovereign AI and is prioritizing practical execution over mere symbolism.
In South Korea, sovereign AI is defined as more than just developing a single proprietary language model. It refers to a country securing capabilities and initiative across the entire AI industry value chain and ecosystem – from semiconductors, power, and computing infrastructure to data centers, cloud, and AI deployment services. So why is South Korea, a key U.S. ally and a tech powerhouse in its own right, now taking this risky and expensive path toward AI autonomy? The answer lies in a high-stakes experiment to carve out a 'Third Way' in an age of China-U.S. AI rivalry.
The Geopolitical Imperative: The Two Waves
The entrenchment of this shift was preceded by a pair of immense shockwaves.
The first wave began in late 2022, when OpenAI's ChatGPT swept the globe. South Korea's tech giants at the time responded swiftly, with NAVER unveiling 'HyperCLOVA X' and LG launching 'EXAONE' to jump into the race to develop their own LLMs. It was at this point that pioneers like Ha began to champion the concept of 'sovereign AI.' Ha, in particular, stressed that to protect national sovereignty, critical infrastructure for diplomacy or defense could not be entrusted to foreign firms. He also argued that it is essential to preserve the nation's cultural and historical context – highlighting cases where early U.S. models referred to the 'East Sea' as the 'Sea of Japan,' a highly sensitive issue in South Korea.
However, this initial push for sovereignty faced three major obstacles. First, the dominant global AI discourse at the time – from the 2023 AI Safety Summit at Bletchley Park in the U.K. to the 2024 AI Seoul Summit – was framed around cooperation over competition. Second, the Korean public's skepticism toward nationalism meant such arguments were often met with domestic derision as 'guk-ppong (국뽕),' a slang term for jingoistic fervor.
Critically, there was widespread skepticism that the concept of sovereign AI itself represented specific corporate interests. As Ha was the AI chief of NAVER at the time, critics dismissed the argument as 'marketing jargon' aligned with the company's commercial interests. This perception was reinforced when global business leaders like Nvidia CEO Jensen Huang also endorsed sovereign AI, due to the pragmatic calculation that chipmakers stood to gain the most as nations rushed to purchase GPUs for their sovereign projects.
What shattered this stalemate and elevated the sovereign AI discourse to the forefront of the nation agenda was the second shockwave: the 'DeepSeek Moment.' The V3 model from the Chinese startup DeepSeek, which gained attention around the Lunar New Year in February 2025, shocked the world with its claim of having lowered training costs to $6 million through algorithmic optimization (though experts point out this claim is debatable).
The key takeaway was that a geopolitical rival, China, could create a frontier AI in a 'very cost-effective' manner. This signaled to South Korean decision-makers that sovereign AI was no longer a game only for the capital-rich United States, but had now become a viable and essential mission for South Korea as well.
The Chief AI Officer's Doctrine: A Vision of 'Full-Stack' Sovereignty
To understand Seoul's strategy, one must first understand the blueprint of its new architect, Chief AI Officer (CAIO) Ha Jung-woo. His argument for sovereign AI goes beyond simple technological nationalism; it is a comprehensive vision rooted in South Korea's significant tech capabilities and the geopolitical realities of the 21st century.
At the core of Ha's argument is the belief that South Korea is one of the few nations with the capacity to even attempt such a feat. He has frequently emphasized that the country possesses nearly all the components of the AI 'full-stack': world-leading memory semiconductors and burgeoning AI chip (NPU) development, established cloud infrastructure, vast high-quality datasets from a hyper-digitalized society, and a significant pool of AI talent. For South Korea, therefore, sovereign AI is not just a vague dream but a viable national strategy.
Ha defines sovereign AI as liberation from technological dependency. According to him, this is not a call for isolationism, but for securing AI systems that deeply understand and align with the nation's own language, culture, laws, and social values. The argument is that no matter how advanced a foreign model may be, it cannot, by definition, internalize this nuanced context.
Ultimately, his vision is geopolitical. He warns that AI is poised to become the primary tool of 21st-century 'neo-imperialism.' Just as past empires used guns and ships to colonize territories, he argues, future powers will use AI to dominate the digital territories and data of other nations, subordinating their industrial ecosystems. From this perspective, sovereign AI is a defensive shield against this tide.
Moreover, this defensive logic leads to a proactive global strategy: the Third Way. He argues that South Korea must make a unique proposal to nations in the Middle East, Southeast Asia, and beyond that are wary of the technological hegemony of both superpowers, the U.S. and China. This goes beyond mere product sales to a partnership that includes co-development, technology transfer, and a commitment to data sovereignty to help those nations build their own sovereign AI. Given its position as a middle power, this would allow Seoul to establish itself not as a tech hegemon like Washington or Beijing, but as a reliable and non-threatening partner for the digital age.
The Blueprint: A State-Led Push
The shockwaves from DeepSeek were so powerful that they prompted immediate action from Seoul, even amid the political vacuum following the martial law crisis of the former Yoon Suk-yeol administration. In January, the government – then under an acting president – officially unveiled its blueprint for AI by announcing two flagship projects: the Independent AI Foundation Model initiative and the establishment of a National AI Computing Center.
The 2.5 trillion won (approximately $1.84 billion) computing center plan, in particular, drew attention as it was announced shortly after the Trump administration's reported $500 billion 'Stargate' data center plan in January. It was an ambitious vision for the state to directly provide the core infrastructure for AI, but it soon collided with the cold realities of the market. Due to uncertainty over profitability and corporate reluctance regarding the government's demand for a 51 percent controlling stake, the National AI Computing Center failed to attract a private partner in two consecutive bidding rounds. The government is now completely re-evaluating its approach to revive this massive project.
In contrast, the Independent AI Foundation Model project, which had been designed throughout the first half of the year under the tentative name 'World Best LLM (WBL),' was officially launched in June. This project, reflecting lessons from the initial setbacks of initiatives like the National AI Computing Center, was designed with more market-oriented criteria and aims to select a national open-source foundation model. With nearly all of South Korea's LLM developers participating, the final output, born from a fierce competitive evaluation, is scheduled to be released as open-source to the entire domestic AI ecosystem.
The Internal Debate: A Pragmatic Clash of Visions
While few in Seoul oppose the core values of AI sovereignty itself, the government's method has ignited a fierce and complex debate, pitting national ambition against market pragmatism. The most vocal critics come from the country's dynamic startup and vertical AI scene.
Their primary concern is the fear of a new tech chaebol. With former top executives from the nation's two largest tech companies now at the helm of national AI policy, many startup founders worry the strategy will inevitably favor the two giants they hail from. A prominent Korean AI startup CEO wrote on social media, that 'it will create a very strange structure where foreign AI giants, domestic MSPs, and countless B2B software partners will have to band together to compete against the government's AI.' His sentiment was echoed by others who fear a tilted playing field where national resources are funneled to large corporations.
This fear is rooted in a fundamentally different business reality. For most startups, the race to build the biggest foundation models is already a lost cause – a money pit that they believe South Korea cannot afford. The real market, they argue, is in 'vertical AI': creating specialized applications for industries like medicine, law, and finance. For them, U.S.-based Application Programming Interface (APIs) from OpenAI or Anthropic 'are not a threat,' but a vital, cost-effective tool for innovation and survival.
From this perspective, the government's focus on creating a national LLM is a misguided use of taxpayer money. It diverts resources away from South Korea's most vibrant and profitable AI sector – the application layer – to chase a goal that is both commercially unviable and potentially harmful to the broader ecosystem. While the government counters that a national LLM is a public good – an 'AI highway' for all startups to use – this explanation has done little to assuage the market's skepticism so far.
Conclusion: A Grand Experiment of Opportunity and Risk
South Korea's sovereign AI strategy has become a fascinating experiment of a nation navigating the turbulent waters of the 21st century tech race. It is a situation where a grand, state-led geopolitical vision is in stark confrontation with a pragmatic reality that follows market logic. Despite these internal debates, the true stage for this experiment is already global, and its success depends on how it manages the opportunities and risks inherent in its 'Third Way' vision.
The opportunity is clear. For nations in the Global South wary of both U.S. technological hegemony and Chinese influence, the partnership proposal from a middle power like South Korea is a viable card to play. The strategy of approaching them as a reliable partner rather than a tech hegemon can yield new markets and diplomatic assets. However, the strategy carries inherent risks as it relies heavily on the intangible asset of trust. At the same time, potential partner nations question the competitiveness of South Korea's AI models, and some skeptics suggest that the country could merely be used as a negotiation lever against the United States and China.
Ultimately, the success of this grand experiment hinges on whether South Korea can solve a dual challenge: winning the support of the domestic tech ecosystem internally, while proving its technological credibility externally. Can Seoul bridge this gap? Can the state-backed national champions create a platform for coexistence instead of competing with domestic innovators, and persuade their own ecosystem to endure short-term pain and perceptions of unfairness for long-term strategic gain? The world's middle powers, seeking a path between the U.S. and Chinese spheres of influence, are watching this process closely. Whatever the outcome of Seoul's high-stakes gambit, the journey will offer important lessons on the price of and challenges to pursuing autonomy in the age of AI.