
Sovereign AI is about having a choice, not being tethered to any single country's AI systems.
Developers are creating local models for languages like Cantonese and Bahasa Indonesia.
Costs may be lower with focused ambitions and open-source models.
What happened
Governments and companies outside the US and China are building local AI models to avoid relying on foreign systems. Examples include Votee AI's Cantonese model in Hong Kong, Indosat's Sahabat AI for Bahasa Indonesia, and Humain's Arabic model backed by Saudi Arabia's Public Investment Fund.
Why it matters
These regions worry about being cut off or tied to one country's AI. They aim to protect local data, bolster domestic industry, or secure autonomy, as smaller economies fear losing access to essential AI services.
What to watch
Sovereign AI can be affordable. Votee AI trained its model for about $250,000, far less than the tens of billions spent by major AI developers. Open-source Chinese models provide another cost-effective route.
Ask the AI about this article →
The push for sovereign AI is partly a reaction to the dominance of US and Chinese AI systems, which may not serve languages beyond English and Mandarin. Local startups like Votee AI are stepping in to fill that gap, showing that building native models is feasible, especially when the goal is practical rather than frontier-reaching. The involvement of governments and big investors, such as Saudi Arabia's PIF, lends credibility and resource.
Affordability remains a key factor. Votee AI's modest training cost suggests that sovereignty does not always require massive budgets if the model's scope is restrained. This is further enabled by freely available open-source models, which allow countries to customize AI outputs without starting from scratch.
The true value of these initiatives may lie in preserving local languages and industries, rather than in achieving technological supremacy. While the article frames AI sovereignty as providing choice over dependency, it remains to be seen how these models will perform in practice and whether they can sustainably compete on quality and reliability over time.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
Infosys, a founding partner of CrowdStrike's Project QuiltWorks, is bringing enterprise context to CrowdStrike…
Cisco has announced a target to have zero engineers writing code by the end of October
OpenAI's latest model, Astra, can do more thinking off the scratchpad, according to Transformer

Cecilia Ziniti, former general counsel at Replit, left the company in November 2023 and founded GC AI, an AI s…

Nvidia announced Thursday it has agreed to acquire Hugging Face, a major platform for open-source AI models, d…

Nvidia CEO Jensen Huang developed the world's first NVLink-enabled deep learning system, the DGX-1, a decade a…
