In Davos this winter, tech executives waxed lyrical about the democratising potential of artificial intelligence (AI). But across the world, millions of people worry about how they will pay their children’s school fees, and internet access remains a luxury many simply can’t afford. The distance between these two realities captures the central paradox of our age: as AI promises to unite humanity, it risks creating a divide as wide as any ocean.
The UN warns that 2.6 billion people still lack basic internet access, while AI is projected to become a 4.8 trillion US dollars market by 2033. This isn’t just another digital divide—it’s potentially something far more entrenched. Where previous technologies eventually trickled down to the masses (mobile phones now outnumber people in many developing countries), AI presents steeper barriers to entry. It demands not just connectivity but computational literacy, quality data, and significant processing power.
The result threatens to encode today’s inequalities into tomorrow’s algorithms.
The new mathematics of exclusion
Walk into any corporate boardroom in Manhattan or Mayfair, and you’ll hear breathless talk of AI transformation. The reality is more prosaic. While 78 per cent of companies claim to use AI, the devil lurks in the details. Large corporations are systematically outpacing smaller rivals in meaningful AI adoption.
The concentration of power tells the real story. Just 100 companies—overwhelmingly American and Chinese—control 40 per cent of global AI research and development spending. Meanwhile, 118 countries find themselves entirely excluded from global AI governance discussions. It’s akin to writing international maritime law while ignoring every landlocked nation.
For individual workers, the arithmetic is even less forgiving. Take America, where Black workers are 10 per cent more likely to hold jobs vulnerable to AI automation, yet remain underrepresented in the technical fields creating these systems. Women face similar contradictions—more exposed to AI-driven workplace changes but equipped with fewer relevant skills. The technology meant to augment human capability may instead amplify human disadvantage.
When robots come for the middle class
Previous automation waves targeted blue-collar work. AI takes a different approach, going after cognitive labour—the very jobs that parents urge children to pursue for economic security. Paralegals reviewing contracts, radiologists reading scans, financial analysts crunching numbers: all find their expertise increasingly replicable by algorithms.
The transition isn’t uniformly grim. Most small businesses using AI report it enhances rather than replaces workers, with staff reporting 80 per cent productivity improvements. The crucial difference lies in positioning: are workers AI’s partners or its replacements?
But partnership requires access and understanding. During recent focus groups in London, officials asked participants: “Are you concerned AI will impact your job?” The response was telling: “I don’t know, should I be?” This encapsulates the challenge—millions lack basic awareness of AI’s implications for their livelihoods.
Corporate Darwinism, Silicon Valley style
In business, AI creates new hierarchies based not on traditional advantages but on data quality, computational resources, and technical talent. But while McKinsey’s research reveals a telling pattern (large organisations systematically outperform smaller ones in AI adoption practices), there is a twist.
Distribution resembles a barbell. Large corporations (7.2 per cent adoption rate) and small businesses (5.5 per cent) use AI more than medium-sized companies. The giants afford comprehensive strategies; the minnows benefit from accessible tools. Those in between—lacking resources and agility—get squeezed.
This matters because AI could add trillions of US dollars to annual global output. Benefits will flow disproportionately to early adopters. Companies ignoring AI risk joining the ranks of businesses that dismissed the internet in the 1990s—not immediately obsolete but progressively irrelevant.
The talent shortage exacerbates these dynamics. While 73 per cent of employers prioritise AI hiring, 75 per cent complain of inadequate candidate pools. This bidding war primarily benefits those already working at well-funded firms in major tech hubs.
The geography of algorithmic advantage
Location increasingly determines AI access. Economic projections suggest China will capture seven trillion US dollars in AI-driven GDP gains by 2030, North America 3.7 trillion US dollars. For the rest of the world, including Europe, the projections look less promising.
Infrastructure gaps tell part of the story. Sub-Saharan Africa’s internet penetration sits below 30 per cent, compared to North America’s 80-plus per cent. Without basic connectivity, AI discussions remain academic. Even where internet exists, slow speeds, expensive data, and unreliable power limit meaningful access.
Data representation creates deeper problems. AI systems trained predominantly on Western datasets often fail to reflect Global South realities—languages, cultural contexts, economic patterns. The Global Digital Inclusion Partnership highlights how this creates AI that works brilliantly for Silicon Valley engineers but poorly for Kenyan farmers or Bangladeshi merchants. Try asking ChatGPT some simple questions in Swahili: the answers will be shorter and far less relevant than those for English, French, Mandarin.
Algorithmic bias compounds these issues. The National Institute of Standards and Technology tested 189 facial recognition systems in 2019, finding systematic discrimination against darker-skinned individuals. When such systems influence hiring, lending, or policing decisions, they don’t merely inconvenience—they institutionalise disadvantage.
Building bridges
The AI divide isn’t immutable. Thoughtful programmes worldwide demonstrate that strategic intervention can democratize access.
The World Economic Forum’s EDISON Alliance provides a compelling model. Since launching, it has connected over one billion people to essential digital services across healthcare, education, and finance in more than 100 countries. By coordinating public and private resources, the alliance creates digital infrastructure foundations necessary for AI participation.
American historically black colleges and universities (HBCUs) offer another template. Morehouse College’s ‘AI in Basketball’ course, developed with industry partner Stats Perform, has expanded across HBCUs. Such initiatives don’t merely train future AI practitioners—they ensure diverse perspectives shape AI development from the ground up.
The nonprofit internXL demonstrates scalable solutions, offering free AI certifications and connecting students with industry mentors. These programmes share a crucial insight: bridging the AI divide requires more than access—it demands creating meaningful participation pathways.
Policy responses
Governments grapple with shaping AI development while avoiding past technological transition mistakes. Approaches vary dramatically—from Europe’s comprehensive regulation to China’s state direction to America’s market orientation.
Britain’s AI Opportunities Action Plan illustrates one promising direction. Proposed AI Growth Zones would offer enhanced power access and streamlined planning approvals to accelerate infrastructure development. Combining incentives with clear frameworks could help democratise AI capabilities.
International coordination remains essential. International Labour Organisation recommendations emphasise technology transfer, infrastructure investment, and social dialogue protecting workers’ rights. Yet policy often lags technology, with most Global South nations excluded from governance discussions shaping AI’s future.
Addressing the AI divide requires coordinated action across multiple dimensions. Infrastructure investment remains foundational—a 10 per cent broadband penetration increase can boost developing economy GDP growth by 1.4 per cent, providing platforms for AI applications.
Education must evolve beyond computer literacy toward AI literacy. UNESCO’s call for action emphasises community-driven, culturally relevant programs rather than one-size-fits-all solutions developed in distant laboratories.
AI development itself needs democratisation. This means diverse datasets, inclusive design processes, and governance structures incorporating affected community voices. The technology should uplift rather than entrench disadvantage.
International cooperation must transcend voluntary initiatives toward binding commitments on technology transfer, capacity building, and equitable benefit distribution. Otherwise, AI advantages will concentrate among the already privileged—a profound failure of technology and policy.
Business leaders must recognise that bridging the AI divide serves economic as well as social interests. Markets function best with maximum participation; AI systems improve when reflecting diverse experiences and needs.
High digital stakes
The AI revolution unfolds within existing inequality patterns. Left unchecked, market forces seem likely to amplify rather than diminish disparities. Technology promising to augment human intelligence could instead create an aristocracy of the algorithmically advantaged.
Outcomes aren’t, however, predetermined. Previous technological revolutions—railways, telecommunications, the internet—ultimately proved broadly beneficial despite creating temporary disruptions. The key was deliberate action ensuring benefits reached beyond early adopters to society at large.
AI’s impact depends on how we design, implement, and regulate it. Decisions made in coming years about infrastructure investment, educational priorities, regulatory frameworks, and international cooperation will determine whether AI promotes equality or division.
With AI projected to become a market roughly equivalent to the size of Germany’s economy, the question isn’t whether transformation will occur but whether benefits will spread broadly or concentrate narrowly. This depends on whether leaders choose to build bridges across the AI divide or allow it to calcify into permanent global architecture.
The dream of AI democratising human capability remains achievable. But realisation requires more than algorithms and infrastructure and endless new releases of chat models—it demands wisdom, deliberate action, and commitment to ensuring technological progress serves all humanity, not just those fortunate enough to be algorithmically advantaged.
The choice, for now, remains ours. But that particular window will not stay open for much longer.
Photo: Dreamstime.
