Naver in AI: HyperCLOVA X, Korean Benchmarks & Sovereign AI

Korea's search company, with a conglomerate's footprint

Naver launched its search portal in 1999, went public in 2002 and now runs South Korea's dominant search engine alongside commerce, cloud, fintech and webtoon businesses. Naver Webtoon, spun out of the group, listed on Nasdaq in 2024. Naver Cloud is the arm that builds and sells the company's AI models and infrastructure to enterprise and government customers. Choi Soo-yeon serves as chief executive. [1][2]

HyperCLOVA X and "sovereign AI"

Naver introduced its large language model family, HyperCLOVA X, in August 2023, positioning it as a Korean-language model trained on Korean cultural and linguistic context rather than a repackaged Western model. Naver Cloud reported in February 2024 that HyperCLOVA X outscored GPT-4 on a Korean-language evaluation, a vendor-reported benchmark rather than an independently audited one. [3][4]

In June 2025 Naver released HyperCLOVA X Think, a reasoning-focused model that processes text and visual material, including diagrams from Korean college entrance exams, and outperformed rivals on the KoBALT-700 Korean benchmark despite being smaller than global frontier models. Naver has described the model as a cornerstone of Korea's "sovereign AI" ambitions, the effort by Naver and peers like LG AI Research to keep a domestically built, Korean-language model in the market rather than ceding it entirely to US and Chinese labs. [5]

Atlas interpretation: "Sovereign AI" is Naver's term for itself and its national context, not an independent classification, and the company's own benchmark claims should be read as vendor-reported until an independent evaluation confirms them. What is independently verifiable is the strategic logic: Naver is one of a small number of companies outside the US and China fielding a competitive frontier-adjacent language model, built specifically around a national language most global labs treat as a secondary market. [3][5]

Naver on the timeline

Naver is one of sixteen signatories, alongside labs including OpenAI, Anthropic, Google and Samsung, to the Seoul frontier safety commitments signed at the UK and South Korea's May 2024 AI Seoul Summit, a voluntary pledge to publish safety frameworks with risk thresholds and to hold back models that breach them. Naver's presence on that list, as a Korean company hosting the summit's co-organizing country, reflects both its standing as Korea's leading AI developer and the government's interest in having a domestic signatory alongside the foreign frontier labs. [6]