Yoo Won Kim, Naver Cloud CEO: "Sovereign AI Is an Inevitable Trend Amid US-China Rivalry"

"Open-Source Model Release Aims to Strengthen Korea's Sovereign AI Ecosystem"
"Lightweight Models Preferred for Cost Efficiency"
"Calling Imported Foreign Technology Sovereign AI Is Absurd"

Yoo Won Kim, CEO of Naver Cloud, announced open-source artificial intelligence (AI) models available for commercial use and once again emphasized the necessity of AI sovereignty.

At the Tech Meetup held on the morning of the 23rd at Naver Square Yeoksam in Gangnam-gu, Seoul, Kim stated, "Sovereign AI is not just a marketing term but an inevitable trend of the times," reiterating his point.

Naver Cloud CEO Yoo Won Kim (left) and Chief Technology Officer Nakho Sung are answering questions at the Tech Meetup held on the morning of the 23rd at Naver Square Yeoksam in Gangnam-gu, Seoul. Provided by Naver Cloud

Sovereign AI refers to the ability of a nation or company to independently build AI capabilities by securing its own infrastructure and data. Naver has consistently highlighted the importance of sovereign AI as it develops its own AI models.

Kim further explained, "The reason we are open-sourcing AI models today is to strengthen Korea's sovereign AI ecosystem," adding, "Apart from the United States and China, our country is almost the only one with the capabilities required to build sovereign AI."

At the event, Naver Cloud released three lightweight HyperCLOVA X models as open source: the multimodal 'HyperCLOVA X SEED 3B,' as well as the text-based 'SEED 1.5B' and 'SEED 0.5B.'

Open source refers to making a program's source code publicly available so that anyone can use it for new development. With the open-sourcing of HyperCLOVA X SEED, developers from other companies and research institutions can download the models, modify them as needed, and freely use them for business or research purposes.

However, according to the license regulations, competing companies above a certain size will not be allowed to use these models. Detailed license terms will be announced on April 24, when the HyperCLOVA X models are officially released as open source. The HyperCLOVA X SEED models will be made available through 'Hugging Face,' a global AI open-source platform.

Nakho Sung, Head of Hyperscale AI Technology at Naver Cloud, who presented after Kim, said, "At Naver, we use HyperCLOVA X in about 300 projects annually, and around 40 percent of these use lightweight models with 3 billion parameters or fewer," explaining, "If a small model delivers sufficient performance for a specific task, it is preferred for its cost efficiency."

During the subsequent Q&A session, there were also critical remarks regarding other companies' sovereign AI initiatives. Referring to KT's recent announcement to develop sovereign AI in partnership with Microsoft (MS), Kim stated, "Simply importing foreign technology and rebranding it as sovereign is absurd," and added, "Having a nation's fate determined by the order of the U.S. President does not fit the definition of sovereignty."

Regarding global cloud providers such as Amazon Web Services (AWS), Google Cloud, and MS seeking to expand their presence in Korea's public cloud market by calling for relaxed physical network separation standards, Kim asserted, "Given our country's security and safety, they should meet our standards," and added, "Asking us to lower our standards instead is crossing the line."

Addressing the fact that HyperCLOVA X was not mentioned as a major Korean AI model in the AI Index report published by Stanford University, Sung said, "Our model was omitted, so we filed a claim last year," and explained, "We have been focused on technological advancement, which led us to neglect publishing technical reports. We plan to release a technical report in May." In the report, only LG AI Research's 'Exaone' was listed as a major Korean AI model.

Regarding the release timing of the HyperCLOVA X image generation feature, Sung explained, "We have already secured the necessary technology for image generation, but further refinement is needed," and added, "Once investment in GPU resources is decided, we are ready to launch immediately."