South Korea bets free AI can loosen ChatGPT’s grip

[SEOUL] On Sep 4, the executives of SK Telecom, KT and Kakao stood together at the Korea Press Center in Seoul to announce their plans for the nation’s “AI for All” programme, in which each of South Korea’s 52 million citizens would receive free, unlimited access to generative artificial intelligence (GenAI).

No subscription. No token limit. For most users, the front door will be KakaoTalk, already on 50 million phones. No government has tried an initiative of this scale before.

Abroad, the meme followed within hours: “South Korea – gave its entire population free AI. Me using VPN.”

The joke landed because the idea was genuinely remarkable. Governments build roads. They run hospitals. They subsidise strategic industries. They do not, as a rule, hand out AI like it is a public good.

South Korea is attempting something more unusual than other national technology programmes.

It is trying to turn AI into something closer to public infrastructure, and use that infrastructure to make domestically developed models a daily habit, before foreign ones become too familiar to dislodge.

Notably, South Korea did not simply give away AI. It made a trade.

The three consortia at the centre of the programme – led by SK Telecom, KT and Kakao – are not charities. They comprise the country’s dominant telecommunications and platform companies.

The government offered something concrete in return for their participation: 512 Nvidia B200 chips distributed among the three groups, and a commitment to subsidise nationwide operating costs.

In exchange, the operators agreed to build their services primarily on domestically developed AI models. Developers are required to use home-grown AI models for at least half of their systems.

Rather than generosity, it is industrial policy with a consumer interface.

AI for everyone

Each of the three companies is making a different commercial bet.

SK Telecom is routing the national AI through its existing mobile infrastructure – 22.5 million subscriber lines – and building a phone-call interface designed specifically for older South Koreans who are less comfortable with apps.

A user who calls the AI can request 83 types of official government certificates without opening a single application. SK Telecom has set a target of 10 million monthly active users in 2027.

KT is betting on integration. Its consortium, which includes housing platform Zigbang, shopping comparison site Danawa and search portal Daum, is embedding AI into the services South Koreans already use daily.

A user searching for an apartment could move from discovery to application without leaving a single conversation.

Kakao is betting on the front door it already owns. With KakaoTalk as the entry point, the Kakao consortium argues that distribution is the hardest problem in consumer AI adoption, and that it has already solved it.

The government’s commercial logic is clear enough. South Korea has built its AI ambitions largely on chip exports, with SK Hynix and Samsung supplying the memory that powers global AI infrastructure.

AI for All is the demand side of the same bet: If domestic AI services reach scale, they create a sustained market for South Korean hardware, South Korean models and South Korean data infrastructure.

The condition requiring domestic models to anchor the service is not incidental. It is the point.

Whether the programme can effectively change consumer behaviour is less clear.

An ambitious test

South Korea already has one of the highest AI adoption rates anywhere. A recent government survey found that roughly one in four South Korean GenAI users pays for a subscription.

Those users have built habits around foreign models – primarily ChatGPT – that free domestic alternatives will now need to displace.

Yet, free does not automatically beat better. South Korean Science Minister Bae Kyung-hoon acknowledged this at the Sep 4 launch, saying: “If we fail to create a clear point of differentiation from foreign AI, AI for All will fail.”

That ambition has a concrete next step: By 2027, Seoul plans to expand the programme into personalised agents for tasks such as asset management, learning and retirement planning, for all of which local context is essential.

Beta testing begins in October 2026.

The three consortia have until year-end to show that a government-backed domestic service can do something foreign AI cannot – navigate South Korean public services, connect to government systems and complete tasks that require local language and context.

That is the test. Not whether domestic AI is smarter, but whether it is more useful for the locals.

Singapore has taken a different path. Rather than building a national model, it is working on South-east Asian Languages in One Network, an open-source large language model with regional context. Its latest version is powered by Alibaba’s Qwen.

At the same time, Singapore has used grants and programmes to help local businesses adopt whichever AI works best, including foreign ones.

While other Asian governments have weighed similar choices, none has moved this quickly, at this scale. Seoul is now the test case.

If South Korea can make free domestic AI a daily habit before ChatGPT cements its lead, other capitals in the region will study the playbook.

If the service underwhelms and users quietly return to their subscriptions, that too will be a data point, one that other governments may find equally instructive.

For now, the bet is live. Fifty-two million potential users. Three companies with different theories about how to reach them. A government that needs this to work, and knows it.

The question South Korea is trying to answer is the same one every Asian government will eventually face: Can a domestic AI service earn the daily habit of citizens who already have a foreign alternative they trust?

The harder test will come when the subsidy ends.

Free access creates users; it does not automatically create customers. If Seoul eventually withdraws support, domestic providers will have to prove that the habit outlasts the incentive.

Seoul is spending billions to find out.

The writer, a former investment banker, is a technology journalist who writes about Asian business, technology, innovation and capital markets. He previously worked at GE Capital and Korea Development Bank group.