Why access alone will not close the AI divide

Introduction

If advanced AI eventually becomes a major source of education, expertise, planning help and scientific insight, then the key question is not simply who can log in to a chatbot. The deeper question is who can actually convert AI into real capability.

The history of technology suggests that access alone rarely closes gaps. A school can have computers without producing digital literacy. A village can have internet coverage without gaining high-value online jobs. A company can buy AI software without becoming more productive. The same pattern is emerging with generative AI. People, organisations and countries differ not only in whether they possess AI tools, but in whether they have the language support, connectivity, training, institutional capacity and social trust needed to use those tools well.[UNESCO]unesco.orgAI literacy and the new Digital Divide - A Global Call for…6 Aug 2024 — The rapid advancements in artificial intelligence (AI) h…[World Bank]worldbank.orgdptr2025 ai foundationsWorld BankDigital Progress and Trends Report 2025: AI FoundationsThe World Bank's Digital Progress and Trends 2025 report explores how AI…

This matters for the broader idea of AI-enabled human flourishing. The optimistic vision is that advanced AI could make high-level cognitive assistance widely available, helping billions of people learn faster, solve problems more effectively and participate in scientific, economic and cultural progress. But if the supporting infrastructure for using AI remains uneven, the benefits may concentrate among already advantaged populations while others receive only limited gains. The AI divide may therefore become less about access to software and more about access to the conditions that make software useful.

The layers of unequal AI benefit

Public discussion often treats the AI divide as a simple binary: some people have access to AI and others do not. In practice, the divide operates across several layers.

The first layer is basic access. People need devices, connectivity and affordable services. This remains a significant challenge in many regions, particularly where internet access is unreliable or expensive. The World Bank argues that many lower-income countries still face substantial barriers in digital infrastructure and AI readiness despite growing interest in AI adoption.[World Bank]worldbank.orgdptr2025 ai foundationsWorld BankDigital Progress and Trends Report 2025: AI FoundationsThe World Bank's Digital Progress and Trends 2025 report explores how AI…[World]WikipediaWorldThe world is the totality of entities, the whole of reality, or everything that exists. The nature of the world has been conceptu…

The second layer is usability. Having access to a system does not mean understanding how to use it. Effective AI use often depends on skills such as formulating questions, evaluating outputs, checking sources, identifying mistakes and integrating AI assistance into real workflows. UNESCO has increasingly framed AI literacy as a central challenge, arguing that unequal understanding of AI can create a new form of digital exclusion.[UNESCO]unesco.orgguidance generative ai education and researchGuidance for generative AI in education and research7 Sept 2023 — UNESCO's first global guidance on GenAI in education aims to supp…

The third layer is institutional support. Schools, businesses, hospitals, local governments and community organisations all influence whether AI becomes genuinely useful. A teacher trained to integrate AI into lessons can produce very different outcomes from a school where students are left to experiment without guidance. Likewise, a small business with AI training, management support and reliable digital systems may gain much more from the same tools than a business that lacks those foundations.[UNESCO]unesco.orgAI and technologies in educationUNESCO partners with countries and institutions on ideas, research, and evidence-informed digital l…[UNESCO]unesco.orglaunches global roadmap multilingualism digital eraUnveiled during the solutions session “Safeguarding Languages, Empowering…Read more…

The fourth layer is strategic use. Some individuals and institutions learn how to use AI not merely as a convenience tool but as a force multiplier. They employ it for research, software development, scientific analysis, planning and decision support. As these higher-value uses spread unevenly, the productivity gap between advanced users and basic users may grow substantially. OpenAI’s own analysis has argued that differences in institutional readiness and capability use are becoming as important as differences in formal access.[TechRadar]techradar.comWhile 66% of leaders prioritize AI skills development, only 33% of employees report receiving AI-related training. Misalignment between e…

In this sense, AI may resemble higher education more than electricity. The technology can be widely available while its most transformative benefits remain concentrated among those with the skills and structures needed to exploit it.

Why language gaps could become one of the biggest AI inequalities

One of the most overlooked dimensions of the AI divide is language.

Large language models are trained on enormous collections of text, but those collections are highly uneven across languages. English and a small number of other widely digitised languages dominate the data used to build many frontier systems. Speakers of thousands of lower-resource languages often receive weaker performance, fewer specialised tools and less reliable outputs.[arXiv]arxiv.orgarXiv Learning to Adopt Generative AIarXiv Learning to Adopt Generative AI

This is not merely a cultural issue. It affects practical access to knowledge.

A farmer seeking agricultural advice, a student seeking educational support or a patient seeking health information may receive much better assistance if AI systems operate fluently in their native language and local context. When systems perform poorly in those languages, the quality of cognitive assistance declines precisely where it may be most valuable.

UNESCO and other organisations have increasingly warned that AI could reinforce existing linguistic hierarchies if language diversity is not treated as a major infrastructure challenge. Recent research examining more than 6,000 languages argues that AI resources remain concentrated in a small number of linguistic communities and that disparities may be widening rather than narrowing.[UNESCO]courier.unesco.orgUNESCO CourierAfrican languages, the blind spot of AI4 days ago — This initiative aimed to use AI to translate academic texts into local…

The problem extends beyond translation. AI systems also reflect the assumptions, examples, cultural references and institutional contexts present in their training data. A model that performs well for a lawyer in London or New York may be much less useful for navigating local institutions elsewhere.

As AI becomes more deeply embedded in education, administration and knowledge work, language support may become a form of cognitive infrastructure. Societies whose languages are well represented in AI systems could gain easier access to knowledge and expertise, while poorly represented communities risk a new form of digital marginalisation.[UNESCO Courier]courier.unesco.orgUNESCO CourierAfrican languages, the blind spot of AI4 days ago — This initiative aimed to use AI to translate academic texts into local…

Connectivity still matters more than many AI forecasts assume

Some visions of AI abundance focus on increasingly powerful models while paying less attention to the physical systems that allow people to use them.

Yet advanced AI depends on electricity, telecommunications networks, data centres, cloud services and digital devices. The benefits of frontier AI cannot easily reach communities lacking these foundations.

This creates a familiar pattern in technological history. Innovations often appear globally available in theory but remain unevenly distributed in practice because the supporting infrastructure arrives slowly or inconsistently. The internet itself demonstrated this dynamic. Simply making websites available did not guarantee equal participation in the digital economy.

The World Bank’s work on AI readiness repeatedly emphasises that countries require data infrastructure, computing resources, cloud access, digital skills and governance capacity alongside AI tools themselves. Without these complementary systems, AI adoption tends to remain shallow and fragmented.[World Bank]worldbank.orgdptr2025 ai foundationsWorld BankDigital Progress and Trends Report 2025: AI FoundationsThe World Bank's Digital Progress and Trends 2025 report explores how AI…[World]WikipediaWorldThe world is the totality of entities, the whole of reality, or everything that exists. The nature of the world has been conceptu…

This issue becomes especially important in discussions of long-term AI bloom. If advanced systems eventually help accelerate science, education, medicine and entrepreneurship, then gaps in infrastructure could determine which regions participate in those gains and which primarily consume technologies developed elsewhere.

Why training may matter as much as the models

One of the strongest findings from technology adoption research is that tools rarely create value automatically.

People often need time, experimentation and social support before they discover how a new technology fits into their work. AI appears to follow the same pattern.

Recent research on generative AI adoption identifies both a “learning divide” and a “utility divide”. Some groups learn more slowly that AI can be useful, while others discover valuable applications more quickly. The researchers describe a potential “belief trap” in which people underestimate a tool’s usefulness, avoid using it and therefore never gain the experience that would change their assessment. Training programmes can help break this cycle.[arXiv]arxiv.orgarXiv Learning to Adopt Generative AIarXiv Learning to Adopt Generative AI

This finding challenges a common assumption that AI diffusion will naturally occur once systems become cheap enough. In reality, people often need examples, guidance, institutional encouragement and trusted intermediaries before adoption accelerates.

The pattern appears repeatedly:

• Teachers need support integrating AI into classrooms rather than simply being given access.
• Small businesses often need workflow redesign and staff training before productivity gains appear.
• Public-sector organisations frequently need procurement standards, governance frameworks and technical expertise.
• Workers need opportunities to learn new skills before AI complements rather than replaces parts of their jobs.[UNESCO]unesco.orgGenerative AI: UNESCO study reveals alarming evidence…7 Mar 2024 — A UNESCO study revealed worrying tendencies in Large Language…[TechRadar]techradar.comWhile 66% of leaders prioritize AI skills development, only 33% of employees report receiving AI-related training. Misalignment between e…

This means AI literacy may become a major determinant of future opportunity. The societies that teach people how to collaborate effectively with AI systems could gain advantages that exceed the benefits of merely possessing the software.

Institutions can widen or narrow the divide

The most important differences in AI outcomes may emerge at the level of institutions rather than individuals.

Two schools can use the same AI systems and achieve very different results. One may use AI to personalise learning, provide additional tutoring and reduce administrative burdens on teachers. Another may deploy the same technology poorly, generating confusion, dependency or low-quality educational outcomes.

The same applies to governments, health systems and businesses.

UNESCO’s guidance on AI in education repeatedly emphasises human capacity-building rather than technological deployment alone. The organisation argues that governance, teacher training, curriculum design and ethical safeguards are essential parts of successful AI adoption.[UNESCO]unesco.orgAI and education: Protecting the rights of learnersThe rapid digitalization of education and the development of generative artificial int…

This institutional dimension is especially relevant for the broader bloom thesis. If advanced AI eventually provides extraordinary cognitive leverage, then schools, universities, research institutions, libraries, public services and civic organisations may become the mechanisms through which those capabilities spread.

Historically, many of the largest gains from science and technology came not from inventions themselves but from institutions that helped people use them. Public education systems, agricultural extension services, universities and healthcare networks all served as transmission mechanisms for knowledge.

The same may prove true for AI. A society with strong institutions could turn advanced AI into a widely shared resource. A society with weak institutions might see the same technology reinforce existing inequalities.

Bias, representation and whose knowledge gets amplified

The AI divide is not only about access and skills. It is also about whose knowledge, perspectives and experiences become embedded in the systems themselves.

Researchers and international organisations have repeatedly identified problems with bias in large language models. Studies have found patterns of gender stereotyping, cultural bias and unequal representation across different groups and regions.[UNESCO]articles.unesco.orgAIParis, 7 March 2024 – Ahead of the International Women's Day, a UNESCO study revealed worrying tendencies in Large Language models (LLM…Published: March 2024

For users, this creates a subtler form of inequality.

If AI systems understand some cultures, professions and social contexts far better than others, then the quality of assistance becomes uneven. Some communities may find that AI reliably helps them navigate institutions, generate opportunities and solve practical problems. Others may receive advice that is less accurate, less relevant or less culturally informed.

The risk is that AI becomes a global amplifier of existing informational inequalities. Communities already well represented in digital systems become easier for AI to serve, while communities with limited digital representation remain harder to model and support.

Addressing this challenge requires more than technical fixes. It often depends on local data collection, multilingual development, community participation and institutional investment in underrepresented populations.

The fight over broad access is increasingly a fight over capability

As AI systems become more powerful, the central question may shift from “Who has AI?” to “Who can do meaningful things with AI?”

A country where students, workers, researchers and public servants routinely use AI to solve real problems may gain much more than a country where the same tools exist but remain poorly integrated into education and institutions.

Similarly, two individuals may both have access to the same model while deriving radically different benefits from it. One may use it for learning, planning, coding, research and entrepreneurship. The other may use it only occasionally for simple queries. The difference is not access. It is capability.

This distinction is important for evaluating optimistic visions of AI abundance. The strongest version of the bloom case assumes that advanced cognitive assistance becomes genuinely widespread and useful across humanity. That outcome depends not only on model performance but on language inclusion, digital infrastructure, education systems, institutional quality and social capacity.

The future divide may therefore be less visible than the old digital divide. Instead of separating people who are online from people who are offline, it may separate those who can translate AI into lasting capability from those who cannot.

Whether advanced AI expands human flourishing broadly or concentrates advantages among already capable groups will depend heavily on how societies build that surrounding cognitive infrastructure. Access is the beginning of the story, not the end of it.

Amazon book picks

Further Reading

Books and field guides related to Why access alone will not close the AI divide. Use these as the next step if you want deeper reading beyond the article.

Book

Power and Progress

By Daron Acemoglu, Simon Johnson

Directly addresses why access alone does not guarantee shared benefits.

See on Amazon

Book

The Coming Wave

By Mustafa Suleyman

Explores uneven access and governance of powerful AI systems.

See on Amazon

Book

Poor Economics

By Abhijit Banerjee, Esther Duflo

Explains why interventions need local capacity and institutions to work.

See on Amazon

Book

Factfulness

By Hans Rosling, Ola Rosling et al.

Provides context on global inequality and development capacity.

See on Amazon

Browse more on Amazon:Power and ProgressThe Coming WavePoor Economics

As an Amazon Associate I earn from qualifying purchases.

eBay marketplace picks

Marketplace Samples

Example marketplace items related to this page. Use the search link to explore similar finds on eBay.

Shop location

UsingUSA USA USAUKCanadaAustraliaIreland

Example eBay listing

5 in 1 STEM RC Robot Building Kit 795Pcs Blocks App & Remote Control Bricks

SearcheBay.co.uk: educational robot

Browse similar oneBay.co.uk

Example eBay listing

Interactive Educational Drawing Robot Auto Drawing Robot w/ 100 Cards for Kids

SearcheBay.co.uk: educational robot

Browse similar oneBay.co.uk

Example eBay listing

Makeblock mBot STEM Educational Robot Kit – Bluetooth Version Boxed

SearcheBay.co.uk: educational robot

Browse similar oneBay.co.uk

Example eBay listing

TTS Blue-Bot Educational Bluetooth Robot + USB Cable - Fully Tested

SearcheBay.co.uk: educational robot

Browse similar oneBay.co.uk

Browse more oneBay.co.uk

Example items shown for inspiration; availability and pricing can change. Branchoria may earn a commission if you purchase through outbound eBay links.

Endnotes

1. Source: unesco.org
Link:https://www.unesco.org/ethics-ai/en/articles/ai-literacy-and-new-digital-divide-global-call-action

2. Source: arxiv.org
Title: arXiv Learning to Adopt Generative AI
Link:https://arxiv.org/abs/2410.19806

3. Source: unesco.org
Title: guidance generative ai education and research
Link:https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research

4. Source: unesco.org
Link:https://www.unesco.org/en/digital-education

5. Source: techradar.com
Link:https://www.techradar.com/pro/closing-ai-learning-gaps-between-leaders-and-employees

6. Source: techradar.com
Link:https://www.techradar.com/pro/put-plainly-some-countries-are-already-using-ai-to-solve-harder-problems-and-move-faster-openai-wants-to-make-ai-usage-more-equal-between-countries-but-will-it-actually-work

7. Source: arxiv.org
Title: arXiv Artificial intelligence is creating a new global linguistic hierarchy
Link:https://arxiv.org/abs/2602.12018

Published: February 12, 2026

8. Source: arxiv.org
Link:https://arxiv.org/abs/2505.04531

9. Source: arxiv.org
Title: arXiv AI Diffusion in Low Resource Language Countries
Link:https://arxiv.org/abs/2511.02752

10. Source: unesco.org
Title: launches global roadmap multilingualism digital era
Link:https://www.unesco.org/en/articles/unesco-launches-global-roadmap-multilingualism-digital-era

11. Source: courier.unesco.org
Link:https://courier.unesco.org/en/articles/african-languages-blind-spot-ai

12. Source: unesco.org
Link:https://www.unesco.org/en/articles/generative-ai-unesco-study-reveals-alarming-evidence-regressive-gender-stereotypes

13. Source: unesco.org
Link:https://www.unesco.org/en/articles/ai-and-education-protecting-rights-learners

14. Source: articles.unesco.org
Link:https://articles.unesco.org/sites/default/files/medias/fichiers/2024/07/PR_Generative_AI_UNESCO_study_reveals_alarming_evidence_of_regressive_gender_stereotypes_en.pdf

Published: March 2024

15. Source: unesco.org
Title: what you need know about ai and right education
Link:https://www.unesco.org/en/articles/what-you-need-know-about-ai-and-right-education

16. Source: unesco.org
Link:https://www.unesco.org/en/articles/languages-matter-global-guidance-multilingual-education

17. Source: unesdoc.unesco.org
Link:https://unesdoc.unesco.org/ark%3A/48223/pf0000388971

18. Source: unesdoc.unesco.org
Link:https://unesdoc.unesco.org/ark%3A/48223/pf0000395373.locale%3Den

19. Source: unesco.org
Title: qa how unesco driving digital learning and transformation education
Link:https://www.unesco.org/en/articles/qa-how-unesco-driving-digital-learning-and-transformation-education

20. Source: unesco.org
Title: global roadmap multilingualism
Link:https://www.unesco.org/en/global-roadmap-multilingualism

21. Source: unesco.org
Title: and g20 promote linguistic diversity digital age
Link:https://www.unesco.org/en/articles/unesco-and-g20-promote-linguistic-diversity-digital-age

22. Source: worldbank.org
Title: dptr2025 ai foundations
Link:https://www.worldbank.org/en/publication/dptr2025-ai-foundations

23. Source: worldbank.org
Title: emerging technologies
Link:https://www.worldbank.org/en/topic/digital/brief/emerging-technologies

24. Source: worldbank.org
Link:https://www.worldbank.org/ext/en/topic/digital-and-ai/data-and-ai

25. Source: worldbank.org
Link:https://www.worldbank.org/en/about/unit/unit-dec/research/initiatives

26. Source: openknowledge.worldbank.org
Link:https://openknowledge.worldbank.org/bitstreams/99474f35-6a0c-48c3-9104-7b693e04313e/download

27. Source: worldbank.org
Title: generative ai and jobs in lac
Link:https://www.worldbank.org/en/topic/poverty/publication/generative-ai-and-jobs-in-lac

28. Source: openknowledge.worldbank.org
Link:https://openknowledge.worldbank.org/server/api/core/bitstreams/d2ac1ea9-b70e-4080-b5de-8b31098e992f/content

29. Source: openknowledge.worldbank.org
Link:https://openknowledge.worldbank.org/server/api/core/bitstreams/f2509a0f-7153-4f32-b180-bc11e90c4940/content

30. Source: openknowledge.worldbank.org
Link:https://openknowledge.worldbank.org/entities/publication/4a11a37d-149a-44fb-a941-100065ff5eb8

31. Source: openknowledge.worldbank.org
Link:https://openknowledge.worldbank.org/bitstreams/5e2d0789-43dc-4f7b-9804-90a5fcade533/download

32. Source: openknowledge.worldbank.org
Link:https://openknowledge.worldbank.org/entities/publication/6b514a80-1d5a-5e54-9c8d-b51473775613

33. Source: blogs.worldbank.org
Title: What we re reading about the age of AI jobs and inequality
Link:https://blogs.worldbank.org/en/jobs/What-we-re-reading-about-the-age-of-AI-jobs-and-inequality

Published: May 7, 2024

34. Source: documents1.worldbank.org
Link:https://documents1.worldbank.org/curated/en/099123512172521590/pdf/IDU-393e0705-dc57-406a-b146-6c04a2656a43.pdf

35. Source: worldbank.org
Link:https://www.worldbank.org/ext/en/topic/digital-and-ai

36. Source: openknowledge.worldbank.org
Title: It discusses the demand and supply of digital skills.Read more
Link:https://openknowledge.worldbank.org/entities/publication/18cfa834-c2bf-5286-9510-6dc615e96faf

37. Source: linkedin.com
Link:https://www.linkedin.com/posts/welker_protecting-teacher-agency-in-the-age-of-ai-activity-7369676563895730177-Epq1

38. Source: linkedin.com
Link:https://www.linkedin.com/posts/welker_unesco-governance-of-digital-platforms-and-activity-7383863780478263297-WAUg

39. Source: linkedin.com
Link:https://www.linkedin.com/posts/ifc-education_devising-a-strategic-approach-to-ai-activity-7368655733782888449-QOOi

40. Source: linkedin.com
Link:https://www.linkedin.com/posts/cija-augustus-89811a13_globaleducation-educationequity-digitaldivide-activity-7378191151079518208-3Dk6

41. Source: linkedin.com
Link:https://www.linkedin.com/posts/karina-kesserwan-2065266_ai-indigenoussovereignty-dataethics-activity-7302726351755640833–8YG

42. Source: Wikipedia
Link:https://en.wikipedia.org/wiki/World

43. Source: cdut.edu.cn
Title: protecting the rights of learners
Link:https://www.cdut.edu.cn/__local/0/68/31/67B2897D8ADD86271124C1D7A58_F36B9C93_1BFF3E9.pdf

44. Source: fhsmun.org
Title: UNESC O
Link:https://www.fhsmun.org/wp-content/uploads/2025/01/UNESCO-Digital-Equity-in-Education-and-Culture-1.pdf

45. Source: tandfonline.com
Title: The ethics of AI or techno-solutionism?
Link:https://www.tandfonline.com/doi/full/10.1080/01425692.2025.2502808

46. Source: enmaeya.com
Title: 68da3a22f9ce754f0163e8b9 unesco warns ai could deepen inequality in education
Link:https://enmaeya.com/en/news/68da3a22f9ce754f0163e8b9-unesco-warns-ai-could-deepen-inequality-in-education

Additional References

47. Source: oecd.org
Link:https://www.oecd.org/content/dam/oecd/en/publications/reports/2025/12/ai-and-the-global-productivity-divide_f47026c5/c315ea90-en.pdf

48. Source: frontiersin.org
Link:https://www.frontiersin.org/journals/computer-science/articles/10.3389/fcomp.2026.1759027/full

49. Source: sea-vet.net
Link:https://sea-vet.net/images/seb/e-library/doc_file/1417/386693eng.pdf

50. Source: unctad.org
Link:https://unctad.org/system/files/official-document/tir2025ch2_en.pdf

51. Source: linkedin.com
Link:https://www.linkedin.com/posts/the-world-bank-group_future-jobs-activity-7424436592028766208-koLU

52. Source: cecyt17.ipn.mx
Link:https://www.cecyt17.ipn.mx/assets/files/cecyt17/docs/inicio/unescoia.pdf

53. Source: researchgate.net
Link:https://www.researchgate.net/publication/389936132_Integrating_AI_in_Education_Navigating_UNESCO_Global_Guidelines_Emerging_Trends_and_Its_Intersection_with_Sustainable_Development_Goals

54. Source: bera.ac.uk
Link:https://www.bera.ac.uk/blog/digital-equity-in-the-age-of-generative-ai-bridging-the-divide-in-educational-technology

Published: May 2025

55. Source: news.ruforum.org
Link:https://news.ruforum.org/ai-in-education-navigating-fears-embracing-possibilities-and-charting-a-path-forward/

56. Source: linkedin.com
Link:https://www.linkedin.com/posts/jose-f-quesada_lt4all2025-languagetechnology-ai-activity-7293536455686025218-IrtS