Full Replacement by Domestic AI Unlikely,
But Can Serve as a Supplement to Prevent Monopolies
Specialization by Industry and for Korea Needed to Enhance User Convenience

Following the U.S. government’s implementation of export control measures on Anthropic’s latest artificial intelligence (AI) model “Mythos” and the limited release of OpenAI’s new model, GPT-5.6, the need to build independent AI capabilities—known as sovereign AI—is becoming increasingly important. As high-performance AI models are recognized as national strategic assets, the government’s independent AI foundation model project is scheduled to undergo its second phase of evaluation in August. Experts point out that since the ecosystem is already centered around foreign AI, it will be difficult for domestic AI to overturn the market in the short term even if it is developed. However, they believe it could at least play a supplementary role in preventing monopolies.

[Global AI Battleground Korea]⑤A Domestic AI Takeover: Can Korean Models Break the Hold of Foreign AI? View original image

On July 1, Choi Byungho, Professor at Korea University’s Human-Inspired AI Research Institute, said, “Even if sovereign AI is developed, it will be difficult for profit-driven companies that require top performance and service to fully switch to domestic AI due to performance gaps.” He added, “Instead, companies employing multi-model strategies will have more options as the variety of available models increases.”


He further emphasized that the goal of domestic AI should not be broad use as a general-purpose model, but rather securing AI sovereignty and self-reliance. Choi noted that models should be developed to specialize in learning from unique data in sectors such as defense, public services, telecommunications, finance, and energy, where hacking could paralyze national systems or where sensitive data is involved, thereby boosting performance. He predicted that domestic AI would be most useful in business-to-business (B2B) and business-to-government (B2G) environments that require closed work settings and proprietary data.


Professor Choi explained, “Since we cannot compete with the massive infrastructure, funding, and talent pools of the U.S. and China, the goal should not be to increase the performance of general-purpose models.” He continued, “It is necessary to focus research on optimizing AI models for the Korean environment.” He also suggested, “We should pursue multimodal models that can learn from various types of data to develop industry-specific specialized models. By linking this to Korea’s strength in manufacturing technology, it should be possible to transition toward world models specialized in robotics and mobility.”


Lee Jaesung, Professor of AI at Chung-Ang University, also predicted that a single AI model would not dominate the entire ecosystem. Professor Lee said, “Companies will choose the high-performing models, so for the time being, foreign models with superior performance will continue to be used.” However, he added, “In areas such as defense and healthcare, where reliance on foreign models is not an option, domestic models will have to catch up.”


He further stated that in order for domestic AI to become a symbol of AI sovereignty, it must enhance its commercial viability through specialization. Professor Lee said, “We need to target areas where the U.S. and China cannot compete in terms of price or performance. Domestic AI should position itself as an alternative model that is free from U.S. export controls and the risk of Chinese backdoors.”


There was also the opinion that domestic AI will gradually spread and serve as a check on the monopoly of foreign models.


Park Jin-ho, Professor of Computer AI at Dongguk University, said, “Currently, the algorithms of AI models are not significantly different from those overseas, so if infrastructure is built and data is secured as planned, domestic AI will be competitive within two years.” He continued, “Rather than instantly replacing foreign models, domestic AI will gradually expand its scope and help curb the monopolistic dominance of foreign models.”



Professor Park explained, “In the future, performance, reliability, and speed differences among AI models are likely to diminish. Competitiveness will depend on how well the service fits the work styles and lives of Korean users.” He added, “If domestic AI enhances application services and localization for each industry, thereby improving user convenience, it can be sufficiently competitive even if the model itself is not superior in performance.”


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing