LG (003550.KS) has begun full-fledged artificial intelligence model cooperation with emerging economies in Central Asia and the Middle East. The strategy targets demand from countries that urgently need "sovereign AI" to avoid dependence on the United States and China but lack their own technology and data, using the expertise LG has built in developing and commercializing its homegrown EXAONE model.
LG AI Research signed a memorandum of understanding with the government of Kazakhstan in mid-September for cooperation in AI, according to industry sources on the 23rd. Zhaslan Madiyev, deputy prime minister and minister of AI and digital development, who oversees Kazakhstan's national AI policy, visited LG AI Research in person and met with Director Hong-lak Lee.
LG AI Research agreed to use EXAONE technology to jointly develop a Kazakh-language model with Kazakh companies and universities, apply it to industry, energy, healthcare and logistics, and cooperate on training talent.
LG AI Research is also pursuing an MOU next month with a sovereign wealth fund in the Middle East for cooperation on its EXAONE Business Intelligence business. EXAONE BI is a finance-focused model that analyzes some 8,000 listed companies. After securing the London Stock Exchange Group as a client in September last year, the company plans to formally launch EXAONE BI in the Middle East following a proof of concept.
This is the first time LG AI Research has pursued joint AI development or commercialization with non-English-speaking emerging economies. The strategy aims to widen overseas opportunities by targeting a niche where Korean companies can compete on a global AI stage dominated by U.S. and Chinese Big Tech models.
Emerging economies share strong demand for sovereign AI that is tailored to their own languages, cultures, laws and regulations, as in Korea, and that carries no risk of data leaking abroad. But they lack the technical capability to develop and commercialize models on their own. Adopting existing Big Tech models as they are raises concerns about technological dependence on major powers. Those models also fail to properly learn the languages of emerging economies, leaving them less efficient.
According to the World Bank's 2026 World Development Report, English accounted for 61.1% of training datasets by language registered this year on Hugging Face, the AI developer community, followed by other European languages at 26.3% and Chinese at 5.4%. Arabic came to 2.4%, while Kazakh did not register in the statistics.
Korean also fares poorly at 2.2% of AI training data, but Korean companies that have developed their own models and acquired sovereign AI technology and know-how have emerged as key partners for emerging economies. LG AI Research, having recently upgraded EXAONE and solidified its business base at home, is likewise using emerging economies in Central Asia and the Middle East as a bridgehead for overseas expansion.
In April, LG AI Research unveiled EXAONE 4.5, which outperformed OpenAI's GPT-5 mini and Anthropic's Claude Sonnet 4.5 on science, technology, engineering and mathematics benchmarks. It has since developed EXAONE BI, the manufacturing-focused EXAONE Tabula and EXAONE Discovery for new materials research. In line with the group's alliance with Nvidia, it is also jointly developing a next-generation model using Nemotron, Nvidia's AI development platform.
The institute is also upgrading K-EXAONE through a government-led project to build an independent AI foundation model as the country's flagship sovereign AI.
Other domestic sovereign AI developers are also racing to expand into emerging economies, with NAVER developing Thai- and Arabic-language large language models and SK Telecom and Upstage forging similar partnerships.
