Hi everyone! 👋 It’s been a while since I last greeted you, and now that autumn is here, I’m reaching out again.
Today, we're going to elaborate on a specific project we undertook: the AI Training Data Construction Project, which was commissioned by a Korean government agency. We’ll focus on how this work connects with the current hot topic in the industry: Sovereign AI.
The project itself was dedicated to building AI training data from video elements that capture unique Korean cultural sentiment. This included traditional Hanok (Korean house) scenery from K-dramas, the lettering on street signs, and landscapes changing with the four seasons. To properly process this kind of data, we need more than just an AI that can recognize objects; we need one that can truly understand the cultural context, linguistic characteristics, and subtle nuances of color.
Projects like this aren't just necessary in Korea; they are a critical undertaking for every country in the rapidly expanding global AI market.
So, what exactly is Sovereign AI? Let’s explore the concept first.
🤖 What is Sovereign AI?
Sovereign AI refers to a nation directly building and managing its own data, AI models, and AI systems. In short, it's the concept of establishing an independent AI ecosystem without reliance on external entities.
The concept of Sovereign AI emerged within the movement emphasizing data sovereignty. As the importance of personal data protection and national data management grew, the strategy expanded to include the discussion of AI sovereignty. Furthermore, with the rapid pace of AI technological development, this strategy has now broadened to encompass the establishment of self-reliant systems that are protected and fostered at the national level.
To summarize, Sovereign AI is a concept that embodies the direction: 'We will build our nation's AI using our own data and infrastructure, according to our laws and standards, to benefit our national industrial development.'
🛡️Why is Sovereign AI Necessary?
Sovereign AI is becoming increasingly important, not only in Korea but in many countries worldwide. The reasons can be broadly grouped into four categories: ① Security and Privacy, ② Regulatory Compliance, ③ Strengthening Economic and Technological Self-Reliance, and ④ Cultural Context.
① Security and Privacy
AI is comprised of massive amounts of data. If this data is transferred overseas, there is a risk that sensitive information, such as personal data protection or national secrets, could be exposed externally. Sovereign AI provides a foundation for securely managing data within the country and protecting privacy.
② Regulatory Compliance
Laws and regulations differ from country to country. Freedom of expression, levels of personal data protection, and methods of industrial regulation all apply differently across nations. Relying on external AI systems makes it difficult to reflect the unique specificities of our country. Sovereign AI enables the operation of customized AI that takes our specific laws and regulations into consideration.
③ Strengthening Economic and Technological Self-Reliance
AI is a key technology that will determine national competitiveness in the future. Excessive reliance on a specific company or country risks the subjugation of the entire industry. Fostering our own data infrastructure, models, and technological ecosystem through Sovereign AI is of great significance for strengthening long-term economic and technological self-reliance.
④ Cultural Context
The final important aspect is cultural context. For example, when people hear the word 'palace,' the image that comes to mind can differ greatly around the world. People in Egypt might think of pyramids or temples, people in France of the Palace of Versailles, and people in Korea of East Asian traditional palaces like Gyeongbokgung.
The image below was generated by Google Gemini. It shows the results when the word 'palace' is envisioned by a Korean, a French person, an American, and an Egyptian, respectively. You can see that the difference across cultures is quite significant, can't you?
<AI-Generated Images: Demonstrating the cultural variance in generating an "ancient palace">
However, if we use an AI system built only with data and models from other countries, the results could be completely different from what we expected.
🌎 Global Trends and Recent Developments
Sovereign AI is a rapidly accelerating trend worldwide. Let's look at how major countries are responding.
The European Union (EU) is emphasizing data and AI sovereignty through the GDPR (General Data Protection Regulation) and the AI Act. A sense of crisis regarding falling behind the U.S. and China seems to be fueling these efforts. The core principle is: 'European data must be utilized within Europe, according to European standards.'
China is forcefully pushing a state-led AI strategy. Leveraging the vast amount of domestic data generated by its massive internal market, it is combining this with regulations to build a distinct Chinese-style AI ecosystem.
The United States, home to the world's leading AI technology, sees major Big Tech companies—such as Google, OpenAI, and Microsoft—effectively exercising AI sovereignty. Recently, the government has also been increasing its involvement, emphasizing 'Trustworthy AI' and 'AI for national security purposes.'
Singapore and Middle Eastern countries are making noticeable moves to secure AI infrastructure independently. Their strategy is to build their own data centers rather than rely on global Big Tech companies.
South Korea recently announced a government roadmap aiming to enter the top three AI powerhouses globally. By clearly stating its commitment to promoting Sovereign AI, the country is preparing to foster a corresponding industrial ecosystem.
As mentioned above, Sovereign AI is emerging as an accelerating global trend. Key developments include: ① Localization of AI infrastructure, ② Development of nation-specific language/cultural datasets, ③ Opening of public data, and ④ Regulatory reform.
In short, countries worldwide plan to directly secure AI infrastructure and data and, based on this, cultivate a safe and trustworthy AI ecosystem.
This post introduced the concept, significance, and trends of Sovereign AI. Starting with Part 2, we’ll share our ongoing project to implement Sovereign AI — including early test results and insights.
Thank you for reading. See you in Part 2!
