Policy Brief

AI Access & G20 Economies: Small Steps, Big Impact

Efforts to make artificial intelligence (AI) widely available, like South Korea’s ‘AI for All’ and India’s IndiaAI Mission, show how G20 countries are preparing for an AI-focused future. This commentary examines how the national AI strategies align with each country's technical capabilities, the risk of unequal adoption, and the potential for diverging growth trajectories across the G20.

What Works and What Doesn’t

Studies show that AI can help workers be more productive by supporting their tasks and making information easier to access. Research also links the use of AI in companies to more jobs and higher sales. However, not all firms, industries, or groups adopt AI at the same rate. This means there are big opportunities, but also a lot of uncertainty about who will benefit and how quickly.

South Korea’s “AI for All” is one of the recent government efforts among G20 countries to treat AI access as a public good. The program aims for two-tiered public offerings - an open chatbot and a proactive public-service agent-combined with a domestic-model mandate requiring at least 50% of inference to use sovereign Korean foundation models. This design addresses digital equity and national industrial policy simultaneously, but it reveals two central trade-offs.

First, engineering capacity is a significant factor. The initial allocation of Graphics Processing Unit (GPUs) allows only pilot-scale traffic, which shows that ‘free access’ is as much an operational difficulty as it is a policy position. Second, the requirement for the use of domestic models creates a policy dilemma – it gives priority to local models and it may lead to better performance tailored to the local language and promote the growth of the domestic industry; but it could result in a two-tiered experience if the cutting-edge commercial models are still considerably more capable in certain tasks.

India’s AI policy rhetoric emphasizes equitable access, population-scale skilling, and an open, human-centric approach to AI. India’s comparative advantages which include large-scale digital payments, extensive low-cost computing adoption, and multilingual needs, suggest substantial potential to leverage AI for development objectives. The cost of AI adoption can be lower than in many advanced economies because of India's relatively competitive IT services, engineering talent and expanding digital infrastructure. However, achieving inclusive benefits depends on investments in human capital, local-language models, and robust governance frameworks that ensure safety, transparency, and remedial mechanisms. India’s emphasis on talent mobility and district-level High-Performance Computing (HPC) infrastructure indicates a strategy to scatter capabilities across regions and socio-economic groups, which, if implemented effectively, can mitigate the pattern of uneven adoption seen elsewhere.

Making it Fairer

A common worry is that AI adoption could make inequality worse unless supported by strong policy. In the past, new technologies have often increased inequality when they mainly benefit firms or regions that are already ahead. Evidence also shows differences in adoption by gender and company type. Important policy tools to address these challenges may include targeted reskilling and lifelong learning, public access to high-quality AI services to close access gaps, investments in digital infrastructure, and labor policies to prevent too much concentration, and steps to make models transparent and safe. Composite measures of AI readiness, like the IMF index and cross-country tools that combine investment, patents, infrastructure, and skills, help identify where action is needed most.

Potential Pathways for G20

There are three points to consider here. First, automation through AI may displace routine tasks, leading to shifts in labor demand across occupations. Second, AI’s ability to reduce information barriers can create complementarities, boosting productivity across sectors and potentially generating new tasks and markets. Third, AI-driven innovation in products and services can increase labor needs in certain areas, partially offsetting job displacement. Ultimately, the macroeconomic impact will depend on how these forces balance out and whether wage and income gains are broadly shared.

For G20 countries that want to make the most of AI in an inclusive way, lessons from South Korea and India suggest three main priorities. First, match universal-access goals with realistic infrastructure plans and phased rollouts that improve computing power, speed, and service quality. Second, pair public support with training and digital literacy programs to help vulnerable groups use AI effectively. Third, track how benefits are shared using standard AI-readiness indicators and labor market data, and adjust policies on tax, transfers, education, and competition as AI adoption grows.

Conclusion

‘AI for All’ is an aspirational policy that can accelerate inclusive development if backed by credible capacity-building, governance safeguards, and active redistributional measures. The transition to AI involves significant costs for G20 economies, including investment in digital infrastructure, data centres, energy, cybersecurity and skilled talent. Businesses, particularly SMEs, also face high costs of adopting new technologies and reskilling workers. For emerging economies, these challenges are greater due to limited infrastructure and skills, creating a risk of a widening AI divide. The key challenge is to ensure that AI-driven productivity gains outweigh these transition costs and are broadly shared.