[Rumman Chowdhury] Sovereign AI: A new frontier of national ambition

Korea Herald

16 hours ago

Should AI be a public good, shaped by local values, or a technology controlled by a handful of geographically-limited tech giants? The rise of 'sovereign AI' -- locally developed, government-backed artificial intelligence systems -- reflects a world wrestling with the urgent questions of power, autonomy, and identity in the digital age. This push is not merely about technological prowess; it is a response to deep geopolitical anxieties, economic ambitions, and cultural imperatives.
The geopolitical landscape is shifting rapidly. The US-China rivalry over AI supremacy has made clear that technology is now a front line in global power struggles. Smaller countries, wary of becoming pawns in this contest, are investing heavily in their own AI capabilities. Brazil and India, for example, have collectively committed billions to build sovereign AI infrastructures, fearing that reliance on foreign technology could compromise their security and autonomy. Across Southeast Asia and the Middle East, national strategies increasingly prioritize AI sovereignty, not just to keep pace with global powers, but to avoid being left behind in a world embroiled in an AI arms race.
Other Asian nations have emerged as major players in sovereign AI, with some viewing them as a 'buffer' between global hegemonic powers China and US. South Korea is aggressively pursuing sovereign AI to establish technological autonomy and cultural relevance, anchored by a landmark $735 billion investment pledge from President Lee Jae Myung. This strategy centers on developing domestically tailored AI systems trained on Korean language data and historical context, moving beyond reliance on foreign platforms. Key architect Ha Jung-woo, appointed as the nation's first Chief Secretary for AI Future Strategy, advocates for "sovereign AI that has learned from Korea's culture and history," prioritizing infrastructure like 5,000+ GPUs for selected "national AI champions".
Crucial to this vision is the SK Group-Amazon Web Services partnership building South Korea's largest AI data center in Ulsan, alongside legislative foundations like the AI Basic Act (effective January 2026). As the EU's peer in comprehensive AI regulation, this law establishes governance for "high-impact" AI systems while promoting public-private collaboration. Importantly, government-supported AI projects will be open-sourced for broad societal use.
The initiative targets strategic sectors like semiconductors and biotechnology, aligning with national goals to rank among the world's top three AI powers by 2030. This includes expanding GPU capacity 15-fold and achieving 70% industrial AI adoption. Security dimensions are addressed through the National AI Security Consultative Group, though the AI Basic Act explicitly excludes military applications from its scope. While open-source frameworks and regional coalitions offer pragmatic pathways, South Korea's sovereign AI ambitions remain a high-stakes gamble balancing cultural preservation against technical and economic constraints. Structural challenges persist -- limited Korean-language datasets, GPU procurement hurdles, and dependence on foreign infrastructure threaten scalability.
For all sovereign AI developers, the economic stakes are enormous. AI is projected to add trillions to the global economy by the end of the decade. Nations like Saudi Arabia and Malaysia are pouring resources into local AI ecosystems, determined to capture this value for themselves rather than see it flow to Silicon Valley or Shenzhen. The message is clear: countries want to be creators, not just consumers, of the AI revolution. This lesson was hard-learned from the early days of the internet, which was dominated by Silicon-Valley based search engines, shopping websites, and social media platforms. Local leaders now seek to foster homegrown talent, build domestic industries, and ensure that the economic benefits of AI accrue to their own citizens.
Culturally, the need for sovereign AI is even more urgent. Most of the world's leading AI models are trained on English-language data, leaving non-English speakers, and their unique cultural contexts, at a disadvantage. Indonesia's massive Bahasa-speaking population and Saudi Arabia's Arabic-first models are early examples of how countries are tailoring AI to their own languages and traditions. These efforts are not just about better translation or customer service; they are about ensuring that AI reflects and respects local values, histories, and identities.
But building truly sovereign AI is no small feat. It requires not just advanced algorithms, but also robust data infrastructure, massive computing power, local talent and effective governance. The European Union's Gaia-X and India's Bhashini initiative are pioneering efforts to create secure, locally controlled data environments. Denmark is investing in domestic supercomputers, while Singapore and Taiwan are developing their own large language models. In addition, many nations are capitalizing on the Trump administration's purge of scientific talent to draw talent away from the United States. These projects are ambitious, but they are also fraught with complexities.
The technical and economic hurdles are formidable. Training a state-of-the-art AI model can cost hundreds of millions of dollars. This price tag is prohibitive for most nations. Maintaining sovereign AI infrastructure is often more expensive than relying on global cloud providers. And while some countries, like Japan, are pursuing hybrid models that combine local development with international partnerships, others risk creating 'AI islands' isolated from global innovation. Critics note Malaysia's sovereign AI delivers 2.3x slower performance than global counterparts despite heavy investment, highlighting viability concerns.
Ethically, government ownership of AI raises a host of thorny questions. On one hand, public control could ensure that AI serves the common good, free from the profit motives and opaque algorithms of private corporations. On the other hand, it raises the specter of state surveillance and censorship. China's Social Credit System, Hungary's illegal public monitoring of LGBTQ+ allies and the UAE's use of AI to monitor government communications are cautionary tales of how sovereign AI can be weaponized for social control. There are also concerns about bias: nationally curated datasets may reflect the biases of majority populations, leaving marginalized groups behind. In addition, these models may serve as tools for propaganda by current political parties -- similar to how Elon Musk has publicly declared his intent to influence Grok, his AI model.
The path forward is not clear-cut. Absolute AI sovereignty may be an illusion in an interconnected world, but so is complete dependence on foreign technology. Some countries are exploring regional coalitions, like ASEAN's shared AI cloud, or open-source collaborations that balance local needs with global innovation. France, for example, is focusing on sector-specific sovereignty in healthcare, while Indonesia is building on open-source frameworks to create models that are both locally relevant and globally compatible.
Ultimately, the quest for sovereign AI is about more than technology. It is a reflection of a world in flux, where nations are reasserting their identities and interests in the face of rapid change. The challenge is to strike a balance between autonomy and collaboration, innovation and control, national ambition and global responsibility. In the end, the most successful countries may be those that can navigate this delicate equilibrium, building AI systems that are both locally rooted and globally connected.