If you are relying on an external artificial intelligence (AI) model or application programming interface (API), when that API is blocked or the service itself disappears, we cannot even get our hands on it. If you do not know the core technology, applications are difficult, and in the end you become broadly dependent on external technology.
Cha Mi-young, Director General at the Max Planck Institute for Security and Privacy in Germany, said this in a group interview held on the 7th at the Korea Science and Technology Center in Gangnam-gu, Seoul, citing the securing of core technology as a key task for Korea's AI competitiveness.
Director General Cha is the first Korean scientist to receive bachelor's, master's, and doctoral degrees from KAIST and then become a Director General at the Max Planck Institute in Germany. She visited Korea on the occasion of attending the 2026 World Congress of Korean Scientists and Engineers, hosted by the Korea Federation of Science and Technology Societies (KOFST) and the Korean-American Scientists and Engineers Association.
The core technology that Director General Cha describes refers broadly to the fundamental technologies that enable directly building large-scale AI models, boosting performance, and improving them reliably. Model design and training, data processing, evaluation, optimization, safety assurance, and interpretability research are all included. It goes beyond adopting AI services or applying them to specific industries, and means the capacity to understand how AI works and fix it when needed.
She cited recent cases of access blocks to Mythos and Fable in the United States to explain the risks of relying on external models. Director General Cha said, If the core technology is in someone else's hands and there is nothing we can do, you end up in a situation where you just have to wait endlessly.
In this context, Director General Cha, in her keynote speech at the World Congress of Korean Scientists and Engineers, cited "mechanistic interpretability research" as an important recent trend in global AI research. It is research that tracks which parts inside an AI model operate to produce a particular answer or judgment. She explained, It is research that looks inside AI's black box, and if we can identify which variables inside an AI model are responding, it becomes possible to erase or change errors.
Director General Cha also said AI governance is possible only on top of this technical foundation. Merely adopting models built by overseas big tech makes it difficult to fundamentally control AI-made errors, biases, and misinformation. If you do not know how AI is making judgments, you cannot fix it when it goes wrong, and it is hard to trust it socially.
She added that Korea's AI opportunity can also be found here. Director General Cha said, Korea is a country that does AI well. By various measures, its AI research capacity is high and there is a lot of talent, and because it is already a country with much potential, we must never lose the core technology.
However, she also said AI competitiveness is not completed by core technology alone. Because AI is not a technology confined to a specific industry but a foundational technology that affects society as a whole, "AI+X" research that combines it with diverse fields such as law, policy, anthropology, and psychology must be carried out together.
Director General Cha said, For AI+X research, there needs to be a framework where AI researchers and researchers from other fields can meet, and before applying AI to society, we also need safe spaces where we can sufficiently experiment, identify failures and side effects, and make fixes.