Sovereign AI: Are Governments Wasting Billions on Domestic AI Models?

TL;DR:

  • Governments worldwide are investing hundreds of billions in “sovereign AI” to reduce dependency on US and Chinese tech giants
  • Singapore, India, Malaysia and others are building regional language models tailored to local needs
  • Critics argue resources would be better spent on AI regulation rather than competing with established players

Countries from Singapore to India are developing their own large language models as part of a trend called “sovereign AI”, but the strategy has ignited debate over whether middle powers can meaningfully compete with tech giants’ massive investments.

Context and Background

Singapore’s government-funded SEA-LION model converses in 11 south-east Asian languages, whilst India’s IndiaAI Mission has committed approximately £1bn to build national AI systems. These initiatives form part of a global movement where governments are attempting to reduce dependency on US companies like OpenAI and Meta, or Chinese alternatives like DeepSeek.

The investments reflect concerns beyond just technical capability. India’s defence ministry, for example, won’t deploy Chinese AI systems due to potential geopolitical biases in training data, whilst reliance on US systems raises data sovereignty concerns. Many US-built AI models struggle with regional contexts—delivering incomprehensible accents in Indian classrooms or providing hybrid US-Indian legal advice that serves neither jurisdiction properly.

Abhishek Upperwal, founder of Indian developer Soket AI, argues that whilst India cannot match the $300–500bn investments from US companies, targeted expertise and brain power can compensate for the funding gap. His company is developing a smaller national LLM roughly comparable to French company Mistral’s releases.

Looking Forward

Researchers at Cambridge’s Bennett School for Public Policy have proposed “Airbus for AI”—a multinational consortium pooling resources from middle powers including the UK, Canada, Germany, Japan, Singapore and others. The initiative has attracted interest from AI ministers in at least three countries, reflecting growing concerns about overreliance on a single nation’s technology infrastructure.

However, critics like AI strategist Tzu Kit Chan argue governments are wasting taxpayer money chasing a rapidly advancing frontier. He advocates for investing in AI safety regulations instead, noting that most Malaysian users still prefer ChatGPT or Gemini over sovereign alternatives. Singapore’s AI Singapore takes a middle path, positioning regional models as complementary tools that encode cultural nuance rather than direct competitors to frontier systems.

Source Attribution:

Share this article