The Sovereign AI Bet: Strategic Imperative or a Billion-Dollar Mistake?
Governments are pouring billions into creating their own "sovereign" AI. The goal is to reduce reliance on US and Chinese tech giants.
But in a market dominated by players with near-limitless funding, is this a strategic move for national security and cultural relevance, or a massive waste of taxpayer money?
The Shortcomings of Foreign AI
Many countries find that AI systems built in the US simply don't work for their needs. An AI teacher in a remote Indian village speaks with a strong US accent that students can't understand.
An Indian legal startup found Meta's LLaMa model produced useless advice, mixing US and Indian law. In Malaysia, models recommend pork-based recipes. These aren't minor glitches; they are fundamental failures of context.
National Security is Non-Negotiable
Beyond cultural misses, there are serious security issues. India's defence ministry considers Chinese models like DeepSeek to be off-limits. The risk of training data containing hostile propaganda-for instance, stating that Ladakh is not part of India-is too high.
Abhishek Upperwal, founder of Indian developer Soket AI, notes that even US systems like OpenAI are a concern. "They don't even want to rely on [US] OpenAI-type systems because data might go outside the country, and that is absolutely not OK with them."
Competing Without Billions
Smaller nations can't win a spending war against the US and China. So, they are adopting different strategies.
In India, the government's IndiaAI Mission has committed $1.25 billion to AI development. Companies like Soket AI aim to build smaller, efficient models, focusing on talent and ingenuity-"the brain game"-to close the funding gap.
Singapore takes a complementary approach. Leslie Teo of AI Singapore says their SEA-LION models, trained in 11 regional languages, are designed to work with larger models, adding the necessary cultural nuance that ChatGPT lacks. The goal is to be "smart consumers" of the technology.
Strength in Numbers
Another option is for countries to team up. Researchers from the Bennett Institute for Public Policy at Cambridge proposed an "Airbus for AI"-a public company funded by a consortium of nations like the UK, Canada, and Japan to create a true competitor to US and Chinese giants.
Joshua Tan, the paper's lead author, says several countries are interested. "There's less trust in the promises of this current US administration. People are asking like, can I still depend on any of this tech? What if they decide to turn it off?"
A Sobering Reality Check
Not everyone is convinced. Tzu Kit Chan, an AI strategist advising the Malaysian government, warns that the pace of AI development is so fast that these sovereign projects risk becoming obsolete before they are even deployed.
"I wish the people who are building these [sovereign] AI models were aware of just how far and just how fast the frontier is moving," he says. "The cost to governments having a bad strategy... is that they waste a ton of money."
Chan suggests a better strategy: spend the money on developing strong AI safety regulations, not on trying to compete with products that have already won the market. After all, he notes, most professionals in Kuala Lumpur are using ChatGPT or Gemini, not a local alternative.
Informed Decisions are Critical
The choice for governments is not simple. Building sovereign AI addresses clear security and cultural needs, but it's an expensive bet against fast-moving, deep-pocketed giants.
Deciding whether to build, buy, or regulate requires a deep and practical understanding of AI capabilities. Government leaders and their teams must be equipped with the right knowledge to avoid costly mistakes and develop a coherent national strategy.
Developing this expertise is a prerequisite for effective policy. For officials looking to build this foundational knowledge, specialized training can provide clarity on what's possible and what's practical. You can explore AI courses for various professional roles to make more informed strategic decisions.
