Artificial Intelligence: the next digital divide

The digital divide was once defined by one question: who could get online, and who could not. But as AI becomes embedded in education, employment, public services and everyday life, a new divide is emerging. Tiago Lopes asks – if society is already subsidising connectivity, should governments think about AI access in the same way?

The digital divide has traditionally described the gap between those who can participate in the digital world and those who cannot. For many years, this meant having both a suitable device and an affordable connection, whether through mobile data or home broadband. Without these, households could be excluded from online education, job applications, banking, public services, health information and basic communication.

Across the world, governments have increasingly accepted that connectivity cannot be treated purely as a private consumer service and, in many countries, introduced social-connectivity measures to reduce the gap. Affordable internet access has become a recurring part of digital-inclusion policy, whether through social tariffs, low-income broadband subsidies, universal service obligations, free mobile data schemes, or device-and-connectivity support.

• In the European Union, universal service rules link adequate broadband access to social inclusion and participation in the digital economy.
• In the UK, social tariffs are available on broadband and mobile services for people receiving certain benefits.
• In Canada, the Connecting Families Initiative offers discounted internet services and, in some cases, low-cost devices to eligible households.
• The Internet Brasil programme provides free mobile connectivity to low-income public-school students.
• In the United States, the Affordable Connectivity Program supported more than 23 million households before its broadband discount ended in 2024.

These examples show a common policy principle: in a digital society, connectivity is no longer a luxury, but a condition for participation.

But this also exposes the next gap: if governments have recognised that low-income households may need support to access the internet, the same question now arises for artificial intelligence.

As AI assistants, AI tutors, AI productivity tools and AI-enabled public services become more important, the divide is no longer only about being connected. It is about whether people can access meaningful AI capability once they are connected.

This is where a new policy gap is emerging: while many countries have social-access programmes for internet or mobile connectivity, very few have equivalent measures for AI.

The AI access ladder

AI access is best understood as a ladder, where each level builds on the one below it. If one level is weak, the next becomes harder to benefit from.

Level 1 – Connectivity
AI tools are mostly cloud-based, so users need more than minimal connectivity: they need enough data and a stable connection to use them well. A household may be technically online, but unstable mobile data or limited bandwidth can still make meaningful AI use difficult.

Level 2 – Device access
A mobile phone may be enough for basic AI use, but many educational, professional and administrative tasks are easier on a larger screen, with a keyboard, more storage and greater privacy controls. The device itself shapes the quality of the AI experience.

Level 3 – AI assistant access
The third level is the availability of general-purpose AI assistants such as ChatGPT, Copilot, Gemini, Claude, and similar tools. For most people, this is the first practical encounter with AI. They do not begin with machine learning models or enterprise systems. They begin by asking questions, summarising information, translating text, drafting emails, preparing CVs, or seeking help with everyday tasks.

Level 4 – Quality and integration of access
Free access is not the same as equal access. Paid or institutionally provided tools may offer better models, higher limits, file uploads, voice features, stronger reasoning, privacy controls, and integration into office software, learning platforms, job portals or public-service systems. This creates a divide between basic AI availability and tools that are sufficiently capable, integrated and useful.

Level 5 – AI literacy
AI literacy means knowing how to use AI safely, critically and effectively; to ask useful questions, assess outputs, protect personal information, spot possible hallucinations and bias, and apply AI to real problems. It is an essential step towards inclusion, but it cannot replace access to the tools themselves.

Seen this way, the AI divide is not just about whether someone can open a chatbot. It stretches from connectivity and device access through to AI tool access, integration and literacy. This is why AI inclusion should be understood across multiple levels, rather than as a single question of access.

From AI literacy to AI access

People need more than just technical access to AI; they also need confidence, judgement and practical skills. This is why governments around the world are increasingly investing in AI literacy programmes.

India’s Yuva AI for All, for example, is a free national course designed to introduce citizens, especially young people, to artificial intelligence.

In some cases, countries are going one step further by linking AI literacy with AI tool access:

• Singapore’s Budget 2026 connects AI training with tool access, offering six months of free access to premium AI services for Singaporeans who take selected AI courses.
• Estonia’s AI Leap programme provides students and teachers with free access to leading AI applications and the skills to use them effectively in learning, making it one of the clearest examples of government-backed AI provision, although it is education-based rather than poverty-targeted.

That distinction is important – current AI inclusion policies are mostly universal, school-based or workforce-based. They help citizens learn about AI, but they do not yet create a welfare-style entitlement to AI access for low-income households. In other words, AI literacy is becoming visible in public policy, while AI access as a social entitlement remains underdeveloped.

The policy challenge is therefore broader than just providing training. A person can attend an AI course but still lack the tools, device, data allowance, privacy protections or subscription needed to use AI regularly. This mirrors the older digital divide: teaching digital skills was useful, but it did not replace affordable broadband. Likewise, AI literacy will not replace affordable access to meaningful AI capability.

The next phase of digital inclusion may therefore require recognising AI as a distinct access layer. Social broadband and mobile-access programmes addressed the connectivity gap – AI literacy initiatives address the skills gap. The emerging challenge lies in ensuring access to useful, reliable and sufficiently capable AI tools.

The risk is that AI becomes another premium layer of digital life: available in theory, but materially more useful to those whose schools, employers or households can afford better tools.

This raises difficult policy questions:

• What constitutes basic AI access?
• Should advanced AI capabilities be considered part of digital-inclusion policy?
• Should schools, libraries, community centres or public-service organisations provide access?
• Should low-income households receive support in the same way that some currently receive assistance with broadband and mobile connectivity?

For telecom operators, this matters because digital-inclusion policy may increasingly move beyond basic connectivity and towards the services, platforms and entitlements that sit on top of the network.

Broadband policy has evolved from viewing connectivity as a commercial service to recognising it as a prerequisite for participation in society. AI policy has not yet undergone a similar transition.

The policy gap is clear: society has begun to subsidise connection, but not yet capability. Until governments define what meaningful AI access entails, who should receive support and how access should be provided, the AI divide is likely to deepen the existing digital divide.

As AI becomes part of education, employment, public services and daily life, the question will no longer be simply who is online, but who has access to the tools needed to participate fully.