«Sovereign AI» push grows as nations seek to avoid dependence on foreign models
Governments and companies outside the U.S. and China are building local AI models to protect data, support domestic industry, and guard against being cut off by foreign suppliers. Developers in Hong Kong, Indonesia, South Korea, and Saudi Arabia are leading efforts to serve languages and needs that global tech giants overlook.
The global push for «sovereign AI» is less about building the most powerful systems and more about preserving the freedom to choose where technology comes from, according to executives and developers working outside the United States and China. As artificial intelligence becomes essential infrastructure, governments and companies in Europe, the Middle East, and Asia are increasingly wary of relying on models controlled by foreign suppliers.
«AI has become such an essential need, and so you don’t want to be tethered to anybody else who can turn it off,» said Pak-Sun Ting, founder of Hong Kong-based Votee AI, in an interview with Fortune. Ting is scheduled to speak at the Fortune Leaders Forum in Macau on September 8, where he will join a panel on AI adoption and business strategy.
The meaning of sovereignty varies by region. In Europe, the priority is data protection and keeping personal information within national borders. In the Middle East and much of Asia, the focus is on building local industry and capturing economic benefits from a homegrown AI sector. For smaller economies, sovereignty is about autonomy and avoiding the risk of a foreign provider cutting off access to critical technology.
Votee AI’s main product is a model that operates in Cantonese, the Chinese dialect spoken in Hong Kong and neighboring Guangdong province. Ting noted that the AI revolution has largely developed in English and Mandarin, leaving other languages underserved. «Cantonese is used in education, health care, and police communications. If those don’t get covered, then AI is essentially useless,» he said.
Votee is part of a growing movement of startups and established firms building models for regions that global tech companies have overlooked. Indosat, one of Indonesia’s largest telecom operators, is developing Sahabat AI, a model focused on Indonesian languages including Bahasa. South Korean companies are competing in a government-sponsored contest dubbed the «AI Squid Game» to create the best domestic model. Humain, an AI company backed by Saudi Arabia’s Public Investment Fund, recently unveiled an Arabic-language model built with the help of Chinese developer MiniMax.
These languages represent substantial populations. Korean and Cantonese each have roughly 80 million speakers, while more than 200 million people speak Bahasa Indonesia. Despite their scale, these «low-resource languages» lack the massive text datasets available for English and Mandarin, making model development more challenging.
Cost remains a significant barrier to sovereign AI efforts. Processors are expensive, data centers require heavy investment, and skilled technical talent commands high salaries. Ting pointed out that governments, the most likely customers for sovereign models, do not necessarily need cutting-edge performance. More modest requirements help keep expenses manageable.
Votee trained its model for approximately $250,000, a fraction of the tens of billions of dollars that major AI developers like Anthropic and OpenAI have invested in their systems. Countries and companies can assemble models, semiconductors, and processing capacity from multiple suppliers, then customize the output for local languages and cultural contexts.
The availability of powerful open-source models from Chinese developers is also lowering barriers to entry, allowing organizations to download and run sophisticated systems without paying licensing fees. This trend suggests that «sovereignty» in AI may ultimately be less about total control and more about maintaining viable alternatives.
