‘Sovereign AI’ is really about having a choice: ‘You don’t want to be tethered to anybody else’
Also: All the news and watercooler chat from Fortune.
“Sovereign AI” is a buzzword gaining traction in the global tech sector, particularly outside the U.S. and China. Governments and companies are wary of becoming overly reliant on AI systems controlled by a single country. The concept of sovereignty in AI varies across regions, with Europe focusing on data privacy, the Middle East and Asia aiming to bolster local industries, and smaller economies seeking autonomy to avoid supplier disruptions.
Pak-Sun Ting, the founder of Hong Kong's Votee AI, explained that the company's main product is an AI model designed to operate in Cantonese, the Chinese dialect spoken in Hong Kong and Guangdong province. Ting emphasized that the AI revolution is dominated by English and Mandarin, and without support for Cantonese, AI is essentially useless.
Votee AI is part of a growing trend of developers, startups, and established companies working on AI models tailored for other parts of the world. For instance, Indosat, a major telecom company in Indonesia, is developing Sahabat AI to focus on Indonesian languages, while South Korean firms are participating in a government-backed competition to create the best homegrown AI model.
Humain, an AI company backed by Saudi Arabia's Public Investment Fund, recently unveiled an Arabic-language model developed by Chinese AI company MiniMax. While the languages involved are significant, such as Korean and Cantonese with around 80 million speakers each, and Bahasa Indonesia with over 200 million speakers, these "low-resource" languages lack the extensive text corpora available for English and Mandarin Chinese.
Building sovereign AI involves more than political will; it requires expensive AI processors, data centers, and skilled tech talent. Ting noted that governments, the primary customers for sovereign AI, do not necessarily need the most powerful models. This more modest approach can help reduce costs. For example, Votee trained its AI model for approximately $250,000, a significant investment but far less than the tens of billions of dollars spent by AI giants like Anthropic and OpenAI.
To obtain models, semiconductors, and processing power, countries and companies can source them from various providers and then tailor the AI's output to better suit their needs. This may indicate that the focus of "sovereignty" in AI is not about control, but rather about choice. Fortune's CEO Daily will observe a holiday on Monday, marking the U.S. Labor Day, and will resume publication on Tuesday.
Written by urgent.news from Fortune's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.