South Korea has just done something with artificial intelligence that should make Australian policymakers, business leaders and technologists stop and think.

It plans to give every citizen free access to domestically powered generative AI. The initiative, known as “AI for All”, is intended to provide universal access to general-purpose AI services built substantially on Korean AI models. The longer-term ambition goes considerably further: a personal AI agent for every citizen, capable not simply of answering questions but eventually helping people access government services, make reservations, complete transactions and navigate everyday life.

At first glance, this could be dismissed as a government-funded chatbot. That would miss the point.

What makes South Korea’s approach interesting is the strategy sitting behind it. The government is using universal access to drive AI literacy and adoption, create demand for Korean AI models, strengthen domestic technology capability, develop its AI industry and progressively integrate AI into government services.

There is a sovereign capability strategy, an industrial strategy and a productivity strategy wrapped around a proposition every citizen can understand: give everyone access to AI.

And it raises an interesting question for Australia. Why have two advanced economies facing essentially the same technological disruption reached for such different policy instruments?

Two surprisingly comparable economies

The comparison between Australia and South Korea is more reasonable than we might initially assume.

South Korea has a population of around 52 million compared with approximately 28 million in Australia. Yet the two economies are remarkably similar in nominal size, at around US$1.9 trillion each.

They are both wealthy democracies, members of the OECD, major US allies and highly connected to the global economy. Both have sophisticated education systems, advanced research institutions and populations that have rapidly adopted digital technology.

They also face some remarkably similar strategic questions. How do they improve productivity as their populations age? How do they maintain competitiveness in an increasingly technology-intensive global economy? How do they manage dependence on foreign technology? How do they develop the skills needed for an AI-enabled economy? And how do they ensure the benefits of AI are broadly distributed rather than concentrated among a relatively small number of companies and highly skilled workers?

Their responses to AI, however, reveal some important differences.

Korea is putting serious money behind the ambition

South Korea’s “AI for All” initiative sits within a much larger national commitment to AI.

Its national AI budget has increased from KRW 3.0 trillion in 2025 to KRW 9.9 trillion in 2026, more than tripling in a single year. That is around A$10 billion.

The universal AI service will initially be supported with access to 512 Nvidia B200 GPUs, with government moving to fund the cost of providing the service nationally from 2027. Participating services are also deliberately designed to make substantial use of Korean-developed AI models rather than simply providing citizens with subscriptions to foreign AI platforms.

The significance of the investment is therefore not simply its size. It is how the different pieces fit together.

Korea wants to develop sovereign foundation models. It is investing in compute and AI infrastructure. It is building skills and AI literacy. It is putting AI into government. It is supporting domestic AI companies. And now it is creating a national user base for those capabilities.

This creates the potential for a powerful cycle. Universal access drives usage, which develops AI literacy and encourages adoption. Adoption creates demand for AI services, which supports domestic models and companies. Those companies create further demand for compute, data centres, engineering capability and skills. Government itself becomes both a major customer and a deployment environment.

In other words, Korea is working on the supply side and demand side of AI simultaneously.

That is an important distinction.

Australia is not standing still

It would be easy, but wrong, to frame this as an ambitious Korea versus a risk-averse Australia.

Australia’s National AI Plan is broader than that. The Australian Government has established three objectives: capture the opportunity, spread the benefits and keep Australians safe. The Plan includes attracting investment in AI and digital infrastructure, backing domestic AI capability, increasing adoption, developing skills and improving public services.

The Government describes AI as an opportunity to create a more competitive, productive and resilient Australian economy and explicitly recognises the need to develop local capability. There is also substantial work underway across AI infrastructure, skills, adoption and the use of AI within the Australian Public Service.

Australia therefore has an AI strategy that extends well beyond regulation and safety.

The more interesting comparison is that Australia and Korea appear to start from somewhat different assumptions about the role government should play in creating technological capability.

Australia’s approach largely asks how we create the conditions for AI investment, innovation and adoption while ensuring Australians are protected from its risks.

South Korea appears to be asking an additional question: how do we make AI a basic national capability available to everyone?

Neither starting point is inherently right or wrong. But they can lead governments towards very different policy instruments.

And the interesting question is why.

Are Korean policymakers simply more technologically literate?

One possible explanation is that Korea has more technologically sophisticated politicians or public servants.

It is tempting, particularly when looking at the prominence afforded to science and technology within the Korean Government. But there isn’t strong evidence to support such a simple conclusion.

Australia performs extremely well internationally in digital government. The OECD’s 2025 Digital Government Index gives South Korea a composite score of around 0.95 and Australia around 0.88, against an OECD average of approximately 0.70. Australia performs particularly strongly in areas including digital-by-design government, user-centred services and government platforms.

These are not the results of a technologically incapable public sector.

Korea nevertheless performs exceptionally strongly across the index, particularly in areas such as being data-driven and proactively using digital technologies to anticipate and respond to citizen needs.

The difference, therefore, may not simply be technical literacy.

It may be something deeper: the institutional role that technology occupies within government and economic policy.

Technology as national economic strategy

South Korea has spent decades treating technological and industrial capability as central to national economic development.

The transformation of South Korea since the 1960s is extraordinary. It moved from a relatively poor economy to one of the world’s leading industrial and technology nations. Korean companies became global competitors in semiconductors, telecommunications, consumer electronics, batteries, automotive manufacturing, shipbuilding and advanced industrial systems.

That history matters because it shapes institutional instincts.

When a strategically important new technology emerges, Korean policymakers are accustomed to asking questions such as: what capabilities must Korea possess? What infrastructure is required? Where are we dependent on foreign suppliers? Which domestic industries could become globally competitive? What role should government investment and procurement play? How do we create the skills, demand and markets necessary to support those industries?

AI therefore fits relatively naturally into an established model of national economic development.

There are visible institutional manifestations of this. South Korea has elevated its Ministry of Science and ICT to deputy-prime-minister level. Its national agenda explicitly includes becoming one of the world’s top three AI powers, becoming the world’s most AI-literate country and creating what it calls a “Universal Basic AI Society”.

Korea’s investment in research and development also provides useful context. It invests around 5% of GDP in R&D, among the highest levels in the OECD. Australia invests closer to 1.7%.

The difference is not simply an AI policy choice. It reflects decades of different choices about the role of science, engineering, technology, research and industrial capability in economic strategy.

Seen in this context, AI for All has not appeared from nowhere.

Perhaps the most interesting question is not why Korea decided to provide universal access to AI. It is what characteristics of the Korean system made such an idea seem like a natural policy instrument in the first place.

AI as national infrastructure

This leads to perhaps the most provocative idea behind the Korean approach.

For most of the modern era, governments have thought about enabling infrastructure in relatively familiar terms: roads, railways, electricity, telecommunications and, more recently, broadband and digital infrastructure.

But what if access to intelligence itself is becoming a form of enabling infrastructure?

That may sound grandiose until we consider what generative AI is increasingly capable of doing. It can help someone write, analyse, translate, research, code, learn, design, plan and solve problems. As agentic AI develops, these systems will increasingly be able to act as well as advise, navigating services, completing forms, interacting with systems and executing tasks.

Access to sufficiently capable AI could therefore materially affect the productivity of an individual worker, student, small business or community organisation.

This creates an important policy question because access to the most capable AI systems is not necessarily evenly distributed.

Large companies can purchase enterprise AI platforms, employ specialists, build proprietary systems and integrate AI into workflows. A small business with five employees may struggle to do any of those things. The same applies to individuals, students, community organisations and not-for-profits.

Universal access changes that equation.

It treats AI capability less like premium software and more like enabling digital infrastructure.

The productivity question

This is particularly relevant to Australia.

Australia has spent years debating its productivity challenge. AI is frequently identified as one of the technologies capable of materially improving productivity, but productivity gains do not occur simply because a technology exists.

They occur when people and organisations learn to use it.

This is why the Korean emphasis on AI literacy is important. Giving people access to AI does not automatically make them productive users. They need to experiment with it, understand what it can do, learn where it fails, develop judgement about when to trust it and change the way they work.

That takes time, but more importantly it takes exposure.

Korea’s approach potentially creates tens of millions of people who are not simply aware of AI but regularly interacting with it and discovering where it can add value.

Technological capability is often developed through use. The same was true of personal computers, the internet, smartphones and cloud computing. AI literacy is unlikely to emerge principally through training courses. Much of it will develop through millions of people solving real problems with AI every day.

Universal access could therefore be viewed not simply as a social program, but as a national productivity intervention.

It also has the potential to democratise some of the productivity gains from AI. If advanced AI capabilities become available only to large organisations and relatively affluent individuals, the technology could widen existing productivity gaps. Universal access provides one possible way of reducing that risk.

Creating a market, not just supporting an industry

There is another important feature of the Korean approach.

Governments frequently support emerging industries through research funding, grants, tax incentives, infrastructure investment and skills programs. These can all be valuable, but they principally operate on the supply side.

The Korean model adds demand.

Giving citizens universal access creates users. Those users generate demand for AI services, which creates scale for domestic providers. Usage generates feedback, experience and data about how people actually use AI. That can improve products and services, which in turn increases adoption.

This is industrial policy operating through the creation of a market.

Government procurement can play a similar role. Governments are enormous purchasers of technology. They can use procurement simply to buy services at the lowest acceptable cost, or they can sometimes use it strategically to help develop markets and capabilities.

Korea appears relatively comfortable with the latter approach.

That does not mean Australia should adopt protectionist procurement policies or attempt to manufacture an artificial domestic AI industry. But it raises a legitimate question about whether public procurement, public services and government-created demand could play a more deliberate role in developing Australian technological capability.

We already accept versions of this argument in defence, infrastructure, energy and advanced manufacturing.

AI may deserve similar consideration.

Sovereign capability and sovereign agency

There is another layer to Korea’s approach that deserves attention.

The universal AI services are intended to rely substantially on domestic models. South Korea is therefore not simply paying for millions of citizens to use foreign AI platforms. It is deliberately using domestic demand to support Korean AI capability.

This creates a direct link between adoption and sovereignty.

The term “sovereign AI” can sometimes be interpreted too narrowly, as though every country must develop its own frontier foundation model, manufacture its own GPUs and build an entirely self-contained technology stack.

That is neither realistic nor necessary for most countries, including Australia.

A more useful concept is sovereign agency.

The important question is not whether a country owns every component of the AI technology stack. It is whether it retains sufficient capability to make meaningful choices as technology, markets and geopolitical circumstances change.

For Australia, that means asking:

• Can we access sufficient compute when we need it?
• Can we protect and make effective use of strategically important Australian data?
• Do we possess enough technical expertise to understand, adapt, operate and govern critical AI systems?
• Can government and industry continue operating if access to a particular foreign technology becomes constrained?
• Do we have trusted international partners who can provide capabilities Australia cannot realistically develop domestically?
• Do we retain enough engineering and industrial capability to respond when circumstances change?

Seen this way, sovereign AI does not require Australia to manufacture every GPU, operate an Australian hyperscaler or build a frontier foundation model entirely by itself.

It requires Australia to make deliberate choices about what we must own, what we must control, what we must understand deeply and where trusted international partnerships provide an acceptable level of sovereign agency.

Korea’s approach is interesting because universal AI access is not separate from this capability-building agenda. By creating domestic demand for Korean models, infrastructure and services, the government is helping to create the market that sustains those capabilities.

Risk and capability are not competing objectives

None of this diminishes the importance of AI safety, governance and regulation.

Powerful technologies create real risks. Privacy, cybersecurity, misinformation, bias, accountability, intellectual property and the reliability of AI-generated decisions all require serious attention.

In safety-critical areas such as engineering, health, transport, energy and defence, professional judgement and governance become even more important as AI capability increases.

But there is a danger if the national AI challenge becomes framed primarily around risk.

Regulation can determine how technology should be used. It cannot create the underlying capability required to use it well.

A country could conceivably develop excellent AI regulation while remaining almost entirely dependent on foreign companies for models, compute, platforms and expertise. That might represent well-governed dependence, but it would remain dependence.

Capability and governance therefore need to develop together.

The objective should not be maximum adoption regardless of risk, nor maximum protection regardless of opportunity. It should be the capability to use AI productively and safely, while retaining sufficient national agency to preserve Australia’s choices.

What could an Australian version look like?

There is no obvious reason Australia could not explore elements of the Korean approach, although simply copying Korean policy would make little sense. Our economies, industrial structures and institutions are different.

The more useful exercise is to ask what universal or broadly available AI capability might mean in an Australian context.

It could mean:

• a secure and capable AI assistant available to every Australian;
• high-quality AI learning support available to students regardless of their circumstances;
• advanced AI capabilities available to small businesses that could never justify building them independently;
• AI tools available to community organisations and not-for-profits; and
• personal AI agents capable of helping Australians navigate increasingly complex government services.

There would be substantial questions to resolve around cost, privacy, security, competition, procurement, data governance and the role of Australian versus international models.

But those are questions about how such an approach might work, rather than reasons not to consider it.

The more interesting policy question is what Australia could build around the resulting demand.

Universal or widespread access could potentially support Australian AI infrastructure, research, skills and companies while giving millions of Australians practical experience using AI. It could create new opportunities for Australian businesses to build applications and services around common AI infrastructure. Government services could become both easier to access and a significant environment for responsible AI deployment.

In that sense, a social policy, productivity policy, digital-government policy and sovereign-capability policy begin to reinforce one another.

Perhaps the biggest difference is institutional instinct

This brings us back to the original question.

Why has South Korea reached for universal AI access while Australia has not?

It probably isn’t because one country’s policymakers understand AI and the other’s do not. Australia’s strong performance in international measures of digital government argues against that conclusion.

The difference may instead lie in institutional history and the instincts that history creates.

Australia has traditionally relied heavily on markets, competition, private investment and international technology providers to deliver new technologies. Government creates frameworks, regulates markets, supports research, provides incentives and intervenes where there is a recognised market failure.

Korea’s economic development created a somewhat different institutional instinct. Government has historically been more comfortable playing an active role in identifying strategic technologies, coordinating investment, developing industrial capability and creating demand.

Those traditions inevitably influence how governments respond to AI.

Faced with the same technological revolution, one system instinctively asks how to create the right environment for investment and adoption. The other appears more willing to ask what national capability it wants to create and then use the resources of government to help construct the market around it.

The interesting question is whether AI is sufficiently important that Australia should borrow a little more from the second approach.

Putting technology at the centre of national economic strategy

South Korea’s model won’t necessarily be the right model for Australia. But it demonstrates what a different approach looks like when technology sits at the centre of national economic strategy.

It treats AI not simply as something to adopt and regulate, but as a national capability to build, a productivity tool to put in the hands of citizens and businesses, an opportunity to improve government services and a means of strengthening sovereign agency.

Australia already has many of the ingredients required to pursue an ambitious approach of its own. We have world-class researchers and universities, sophisticated digital infrastructure, significant renewable energy resources, growing data-centre investment, capable engineering and professional services sectors, and close relationships with many of the world’s leading technology companies.

The question is how deliberately we assemble those ingredients.

Perhaps the Korean example gives us an opportunity to reconsider the starting point for the Australian conversation.

Instead of beginning only with the question of how Australia should regulate and encourage the adoption of AI, we might first ask:

What AI capability do we want Australian citizens, businesses and governments to possess?

From there comes a second question:

What infrastructure, skills, markets, institutions and sovereign capabilities do we need to make that possible?

Those questions lead to a somewhat different national conversation about AI.

That may ultimately be the most valuable lesson from South Korea. Not that Australia should copy its model, but that a country of comparable economic scale has looked at the same technological transformation and imagined a very different role for government.

What would Australia’s approach to AI look like if we started from the same premise?