As the gap with cutting-edge artificial intelligence (AI) models from global Big Tech firms like OpenAI, Google, and Anthropic shows little sign of narrowing, Professor Yoonhyung Kim, a world-renowned scholar in natural language processing at the Massachusetts Institute of Technology (MIT) Department of Electrical Engineering and Computer Science (EECS), has stressed that “South Korea must never stop developing its own AI technology.”

In a recent interview with Maeil Business Newspaper, Professor Kim stated, “It is true that catching up with the highest-performing US frontier AI models is extremely difficult, but that doesn’t mean we can afford not to even try.” He added, “The very act of challenging the development of frontier-level large language models (LLMs) can elevate South Korea’s AI technology and talent pool to the next level.”

The discussion around “Sovereign AI”—an independent AI ecosystem built on a nation’s own data and infrastructure without relying on foreign technology—first emerged with the launch of OpenAI’s ChatGPT in 2022. The concept has regained prominence this year following Anthropic’s release of its ultra-high-performance model “Mythos” and the US government’s moves toward export controls on AI models.

Professor Kim pointed out, “As AI models become more advanced, they become directly linked to each nation’s defense and security issues,” adding that South Korea must “prepare for a scenario where the US government restricts foreign access to frontier models.” He reiterated, “This is an area where we must take on the challenge, even if we fail.”

He argued that securing proprietary technology is especially critical in the AI agent environment. “Not every agent requires frontier-level performance,” he explained. “What’s important is owning your own models tailored to each agent.” In the agent era, where speed and efficiency are paramount, the technical capability to design small, optimized models for specific services becomes more crucial than relying on a single, massive model.

Professor Kim has been collaborating with Naver since 2022 on research to advance AI search technology. Through “Naver Search US,” Naver’s AI center in the United States, the ultimate goal is to develop an agent that not only optimizes search results but also seamlessly connects Naver’s diverse services, including shopping and maps. “We are conducting preliminary research to realize ‘NAVER is all you need,’ where everything a user wants can be resolved within Naver,” he said.

Professor Kim is particularly focused on a “health agent” targeted for launch within the year. “In the health domain, even a single hallucination can lead to serious problems, so we are deeply contemplating how to design the verification layer,” he said.

“FlowBot,” a paper co-authored by Professor Kim and Naver Search US that was accepted at the International Conference on Machine Learning (ICML) 2026, is another achievement aimed at solving the challenges of agent design. FlowBot is an engineering technology that allows AI to autonomously discover and learn the optimal task sequence when multiple agents work together, eliminating the need for manual workflow design by humans. “Naver has various agents for search, shopping, and more, and manually creating workflows for each domain is very costly,” Professor Kim explained. “FlowBot is a concept that allows the agent’s workflow itself to be learned, much like machine learning.”

Regarding Naver’s differentiation, he assessed, “While Google’s main focus is search, Naver is a portal with diverse domains like shopping and places, so the range of problems it can solve within its own platform is much broader.”

As a researcher, Professor Kim is also setting his sights on the next evolution beyond the Transformer architecture, the current universal structure of LLMs. “For a decade, AI has advanced based on the Transformer, but I don’t believe it is the end of architecture,” he said. “Moving forward, next-generation architecture research will be my primary focus.”

On the direction of education in the AI era, Professor Kim advised that “fundamental capabilities like structural thinking will become even more important.” He emphasized, “To utilize AI effectively, the ability to structure your own arguments is essential. Education must help people internalize an understanding of how AI works, while also distinguishing between tasks to delegate to AI and areas requiring human judgment.”