South Korea Opens Bidding for a Free National AI Chatbot to Cut ChatGPT and Claude Dependence

South Korea’s Ministry of Science and ICT opened bidding on July 13, 2026 for a program called “AI for Everyone,” a government-funded, free, unlimited AI chatbot that has to run mostly on Korean-made models. The pitch is blunt: 23 million South Koreans already pay for or use ChatGPT, and the government wants a domestic alternative good enough to pull them off it.

Two or three private companies will be picked to build and run the service. The contract requires at least half of all user queries to route through the operator’s own Korean foundation model, with another 30% coming from other domestic AI companies. Foreign models get a narrow leftover slice, and any cost from using them falls on the operator, not the state.

TL;DR

South Korea’s Ministry of Science and ICT opened bidding on July 13, 2026 for “AI for Everyone,” a free, government-backed AI chatbot meant to reduce the country’s dependence on ChatGPT and Claude. Winning operators must route at least 80% of queries through domestic models built by consortia led by Naver, LG AI Research, SK Telecom, Upstage, and NC AI, and the government will supply up to 512 Nvidia B200 GPUs to the selected companies. A public beta is due by late September 2026, with full national rollout targeted before the end of the year. Applications close August 11. Officials frame this as a sovereignty move: with roughly 23 million of the country’s 51 million residents already on ChatGPT, per Wiseapp-Retail data, South Korea wants a homegrown option positioned as a public utility rather than a subscription product, and eventually as an export model for other mid-sized countries anxious about depending on a US-China AI duopoly.

What was announced

The Ministry of Science and ICT (MSIT) is running the tender under President Lee Jae-myung’s administration. Deputy Prime Minister and Science Minister Bae Kyung-hoon put it this way: “AI for All is more than just a service; it is the calculator and computer of the AI era.”

Key terms of the bid, drawn from the ministry’s tender and coverage from PYMNTS, The Register, and Digital Trends:

• Free and unlimited. No subscription, no usage cap, for all 52 million residents. If it holds, South Korea becomes the first G20 country to offer every citizen a public AI chatbot.
• Domestic model floor. Winning operators must send at least 50% of queries to their own Korean foundation model and at least 30% more to other domestic models, an 80% combined floor. Foreign models can only fill gaps, and the operator eats that cost.
• GPU support. The government will supply up to 512 Nvidia B200 GPUs to the two or three selected operators. Companies have to match that with their own investment.
• Two-part design. Phase one is a general-purpose chatbot for everyday questions and drafting. Phase two, planned from 2027, is a proactive agent that flags government benefits a citizen qualifies for and helps complete the paperwork, plus personalized help with things like asset management, learning support, and retirement planning.
• Timeline. Applications close August 11, 2026. Public beta launches by late September. Full national service is targeted before the end of 2026.
• Funding horizon. Government support for the program runs through 2030, with the scale of continued funding depending on annual evaluations starting in 2027.

Kakao, Naver, SK Telecom, and LG have all signaled interest in bidding, per The Register. The eligible domestic models sit with five consortia the government has already been backing: Naver Cloud’s HyperCLOVA X, LG AI Research’s EXAONE, SK Telecom’s A.X series, Upstage’s Solar Pro, and NC AI’s Varco.

Why this matters

This isn’t Korea’s first move into homegrown AI. The government has spent the past two years backing five domestic foundation-model consortia, in a program separate from AI for Everyone, that supplies GPU access worth roughly 157.6 billion won ($114 million) plus dataset and talent funding, structured so each team’s total support is worth close to 1 trillion won once follow-on vouchers are counted, according to KED Global. The stated bar for those models is to reach 95% of the performance of systems like ChatGPT, and teams have to open-source more than half of what they build. That program is what makes AI for Everyone possible on paper: the underlying models it has to use already have billions of won and years of runway behind them.

Naver’s own numbers back up the ambition. HyperCLOVA X Think, released in mid-2025, scored 48.9 on the KoBALT-700 Korean-language benchmark, well ahead of EXAONE Deep at 33 and QwQ-32B at 32.4, and led on eight separate Korean-language benchmarks according to Korea JoongAng Daily. Korean models genuinely do better on Korean-specific tasks than general-purpose Western models tuned mostly on English data. That’s the technical argument underneath the political one.

The political argument is the more interesting story. President Lee framed it directly: “We have the U.S. and China bipolar AI systems, and to avoid dependency, we are creating our own sovereign AI.” That’s the “sovereign AI” logic now spreading across mid-sized economies: if the frontier models that matter come from two countries, everyone else is a customer, not a participant, in the technology that increasingly runs public services, education, and information access.

There’s a sharper version of that argument too. The Register points out that Anthropic has, at points, been compelled to restrict Claude access for certain foreign nationals under US policy decisions made without Korean input. A government chatbot serving citizen benefits, tax questions, or education support can’t run on infrastructure that a foreign government’s policy shift might switch off. That’s not a hypothetical for Seoul. It’s the reason “reducing reliance” isn’t just industrial policy, it’s a resilience requirement.

What “reducing reliance” means day to day is still an open question. If AI for Everyone works, government agencies and citizens who can’t or won’t pay for ChatGPT Plus or Claude Pro get a genuinely free alternative for drafting, summarizing, and navigating bureaucracy, with a specific focus on the roughly 32% of the population the ministry considers digitally underserved. If it doesn’t, Korea has spent a few hundred million dollars validating what OpenAI and Anthropic already knew: distribution and habit are harder to dislodge than raw model quality.

Frequently asked questions

When does South Korea’s national AI chatbot launch? The Ministry of Science and ICT opened bidding on July 13, 2026, with applications due August 11. A public beta is planned for late September 2026, and a full national rollout is targeted before the end of the year.

Will the chatbot really be free? Yes, that’s the core design. It’s meant to be free and unlimited for all residents, unlike ChatGPT or Claude subscriptions, and the government is funding operator support through at least 2030 with annual reviews from 2027 onward.

Which Korean AI companies are involved? The government requires operators to route at least 80% of queries through domestic models. The eligible model makers are Naver Cloud (HyperCLOVA X), LG AI Research (EXAONE), SK Telecom (A.X series), Upstage (Solar Pro), and NC AI (Varco). Kakao, Naver, SK Telecom, and LG have expressed interest in bidding to operate the service itself.

Why is South Korea doing this instead of just using ChatGPT or Claude? Officials cite both digital-divide and sovereignty concerns: roughly 23 million of the country’s 51 million residents already use ChatGPT, many paying for it, while lower-income and older residents are left out. President Lee Jae-myung has also framed it as avoiding dependency on the US-China “bipolar” AI landscape, and coverage from The Register notes past instances of Anthropic restricting Claude access for certain foreign nationals under US policy, which underscores the resilience argument for domestic infrastructure.

How does this compare to other countries’ sovereign AI efforts? South Korea’s approach is unusually direct: a free, government-funded consumer chatbot with a hard domestic-model quota, rather than just funding for model development. If it launches on schedule, South Korea would be the first G20 country to offer a free public AI service to its entire population, and officials have talked about the model as something they could eventually market to other mid-sized countries facing the same dependency question.

Primary sources and further reading

• South Korea to launch universal basic AI chatbot - The Register on the tender terms, GPU allocation, and the Anthropic-access angle
• South Korea Turns AI Into a Public Utility - PYMNTS on the two-phase design, minister quote, and digital-divide framing
• South Korea wants to give every citizen free, unlimited access to its own AI chatbot - Digital Trends on the bidding timeline and domestic-model quota
• South Korea promises free homegrown AI chatbot this year - TechXplore, with the 23 million ChatGPT-user figure and minister quote on competing with foreign services
• Korea Pairs AI Law With Free Chatbot as 23 Million Quit Paying for ChatGPT - Tech Times on the AI Basic Act context and adoption numbers
• Naver, LG, SK, NC, Upstage named to build S.Korea’s sovereign AI model to challenge ChatGPT - KED Global on the underlying five-consortia foundation-model program and its funding
• Naver unveils homegrown AI model HyperClova X Think - Korea JoongAng Daily on HyperCLOVA X Think’s Korean-language benchmark results
• South Korea wants every citizen free, unlimited AI access - TheNextWeb on the “first G20 nation” framing and program funding through 2030
• AI for everyone: South Korea launches free AI agent project for all citizens - UPI’s initial report on the July 13 bidding announcement