South Korea Launches Nationwide Free AI Access Program

Recently, South Korea officially launched a national AI initiative: everyone can use AI for free and without limits.

South Korea has a population of 51 million, with 23.45 million using ChatGPT—a U.S. product reaching nearly half the population.

The South Korean government responded by funding the initiative itself: On August 28, the Ministry of Science and ICT announced that a consortium of SK Telecom, KT, and Kakao has been selected for the "AI for All" project, which will provide free, unlimited AI services to all citizens by the end of the year, with the condition that 80% must run on domestically developed models.

This is the first G20 economy to build AI as a public infrastructure.

AI model capabilities are concentrated in the United States and China, while other countries are consumers.

According to Wiseapp’s April 2026 data, ChatGPT had 23.45 million monthly active users in Korea, Gemini had 8.45 million, and Claude had 2.41 million—the top three were all foreign products.

South Korea has LG EXAONE, SKT A.X K2, and Upstage Solar, but their user bases are several orders of magnitude smaller.

Fastest-growing and most-used mobile apps in Korea

Domestic models need real users to iterate; users won't migrate from ChatGPT to a weaker product, and data continues to accumulate on foreign companies' servers.

This cycle cannot be broken by the market alone.

The South Korean government has chosen a path unprecedented globally: rather than competing head-on in model performance, it is using public funds to control distribution channels.

The AI Basic Law, effective January 2026, established a legal framework for government intervention in the AI industry. The AI Action Plan 2026–2028, finalized in February of the same year, further introduced the concept of an "AI Basic Society," positioning access to AI as a fundamental citizen right on par with education.

Deputy Prime Minister Bae Kyung-hoon called this project "the calculator and computer of the AI era."

The widespread adoption of calculators was driven by bulk purchases by schools and government institutions.

The business logic of "AI for All" can be summarized in one sentence: The government provides funding, computing power, and access to bring 51 million people into China’s domestic AI ecosystem, using real traffic to grow and strengthen local AI models.

This year, the government allocated a total of 512 NVIDIA B200 GPUs to three consortia, with each GPU providing roughly the equivalent of 3–4 H200 GPUs in inference capacity. Starting in 2027, operational costs will be covered by the national budget, and the service will remain free for at least until 2028.

But money and computing power are merely entry barriers; it’s the rules that truly shape the ecosystem: at least 50% of each consortium’s services must run on domestically developed sovereign foundation models, and an additional 30% must run on models from other Korean companies, totaling an 80% domestic minimum.

Foreign models may only be used in scenarios where the Korean model temporarily cannot cover, and the government will not pay for this portion.

This rule requires each consortium to partner with multiple domestic model providers, inadvertently expanding the order pool for South Korea’s entire AI industry.

The main difference among the three alliances lies in their distribution channels.

SK Telecom (SKT) has 22.5 million mobile users and features a "Mobile AI" that can make calls on behalf of users, powered by the 68.8-billion-parameter sovereign model A.X K2.

Kakao has integrated AI into KakaoTalk, the messaging app used by everyone in Korea, allowing users to make appointments, pay bills, and submit applications directly within the chat window.

KT is pursuing a channel alliance strategy, integrating Daum portal and a group of vertical apps into a single chat interface.

All three paths are betting on the same thing: who will be the first to turn AI from a chat tool into a public agent capable of filing taxes, booking medical appointments, and applying for subsidies—things ChatGPT cannot yet do.

The first version launching at the end of the year is a general-purpose chatbot, but the real test won't come until 2027.

The essence of this design is a demand-side industrial policy.

Over the past decade, the mainstream approach taken by countries to support domestic technology has been to subsidize the supply side—funding chip manufacturers to build production lines and providing R&D grants to AI companies.

South Korea has invested in the distribution side: using public funds to create a protected user base, allowing domestic models to gather data and gain iterative experience in real-world scenarios.

In comparison, this one is larger in scale, has a tighter timeline, and demands a higher standard of consumer experience.

The biggest variable is Naver's absence.

Naver is equivalent to Google in Korea, serving as the country's largest search and lifestyle platform with comprehensive entry points including search, maps, shopping, and payments, giving it far superior conditions to become a national AI assistant than any competitor.

But it didn’t even get involved.

https://www.naver.com/

In January this year, Naver Cloud was eliminated from a government-led sovereign AI foundation model project because the visual encoder of its HyperCLOVA X showed a cosine similarity of 99.5% with Alibaba’s Qwen, failing to meet independence requirements—in simple terms, it was accused of secretly repackaging another model. Qwen .

https://www.cryptopolitan.com/korea-naver-ncsoft-ai-contest/

Samsung is also absent.

In summary, the final outcome is that South Korea’s platform with the most user scenarios and its largest hardware manufacturer are not at the table, while the national project is carried out by two telecom providers and one messaging app.

Retention is another tough problem.

The basic version of ChatGPT is free, and Koreans won't stop using it just because there's an additional free alternative.

Wiseapp data shows that ChatGPT had the fastest user growth among all apps in South Korea in the first half of the year, adding 9.86 million new users.

The government can bring users through the door, but whether it can retain them depends on whether domestic agents can do what ChatGPT cannot—functions deeply integrated with local government systems, such as tax filing, medical appointments, and subsidy applications, which foreign products cannot easily replicate in the short term.

This is the most noteworthy investment in the Korean strategy: exchanging the exclusivity of government use cases for user loyalty in domestic AI.

512 B200 GPUs serve 51 million users, compared to the compute power behind ChatGPT’s 900 million weekly users—the scale difference is evident.

Free commitment until 2028; thereafter, the operator will be responsible for its own profits and losses.

If users are already accustomed to free services by then, converting them to paid plans won't be easier than it is today.

The experiment in South Korea addresses a critical question for all countries outside the U.S. and China: When model capabilities fall behind the leading players, can controlling the distribution channel create a window for survival?

After the year-end launch, the first validation metric will be user retention—when free ChatGPT and free domestic AI are side by side, the choices of 51 million Koreans will determine whether this path represents a replicable model or an expensive lesson.

Reference: https://en.yna.co.kr/view/AEN20260828002600320

This article is from the WeChat public account "New Intelligence Yuan" (ID: AI_era), author: ASI Revelation; editor: Marco