The Great Unfinished Project
In 1948, the Universal Declaration of Human Rights proclaimed that "everyone has the right to education." Seventy-eight years later, this right remains unrealized for hundreds of millions.
The progress is real. Primary school enrollment has become nearly universal. Literacy has spread from a minority to a global majority. More humans are educated to higher levels than at any point in history.
But the gaps remain vast. Roughly 250 million children and youth are not in school.¹ Over 600 million who are enrolled don't achieve minimum proficiency in reading and mathematics. The quality of education varies so dramatically that two students can complete the same grade and have learned entirely different amounts.
A child born in Finland receives among the world's best education; a child born in rural Niger may receive almost none. This is not primarily about intelligence or effort—it's about circumstance, resources, and systems.
AI tutoring offers a possibility that has never existed before: world-class educational support available to anyone with a device and connectivity. A student in Bangladesh could access the same AI tutor as a student in Boston. The marginal cost of serving one more student approaches zero.
This chapter examines what global AI education could mean—the potential for unprecedented access, the new inequalities that might emerge, and the policy choices that will determine whether AI narrows or widens the educational divide.
2026 Snapshot — Global Educational Landscape
Access and Enrollment
Primary education: Over 90% enrolled globally; near-universal in most regions.
Secondary education: ~75% enrolled globally; significant regional variation.
Tertiary education: ~40% enrollment in developed countries; much lower in developing nations.
Out-of-school children: ~250 million globally; concentrated in Sub-Saharan Africa, South Asia, and conflict zones.
Gender gap: Largely closed in primary education globally; persists at higher levels in some regions.
Quality and Learning
Learning poverty: Over 50% of children in low and middle-income countries cannot read and understand a simple story by age 10.²
The schooling-learning gap: Many children attend school but don't learn. Enrollment without education.
Quality variation: A student completing 6th grade in one country may have the skills of a 3rd grader in another.
Teacher quality: Dramatic variation in teacher preparation and effectiveness. Many teachers themselves lack basic competencies in the subjects they teach.
Digital Infrastructure
Internet access: ~67% of global population has internet access; vast disparities by country and urban/rural divide.
Device access: Smartphone penetration high even in low-income countries (~80% of adults globally); computers less common.
School connectivity: Many schools, especially in developing regions, lack reliable internet and devices.
COVID acceleration: Pandemic forced rapid adoption of digital learning; revealed both possibilities and gaps.
Current EdTech Reach
Khan Academy: Available in 50+ languages; millions of users globally; free access to content.
Coursera, edX: University courses available globally; certificate programs; limited free access.
YouTube: Vast educational content in many languages; variable quality.
Regional platforms: Local EdTech companies serving specific markets (BYJU'S in India, Ruangguru in Indonesia).
Mobile learning: Feature phone-based learning for markets without smartphones or reliable internet.
Notable Players in Global Education
International Organizations
UNESCO: UN agency for education. Sets global standards and monitors progress. Coordinates Education for All initiatives.
World Bank: Major funder of educational development. "Learning Poverty" metric. Technical assistance and policy guidance.
UNICEF: Focus on children's education, especially in emergency settings. Technology for development programs.
GPE (Global Partnership for Education): Multilateral partnership funding education in developing countries. $5+ billion pledged.
Large-Scale EdTech
BYJU'S: Indian EdTech company; at one point valued at $22 billion; 150 million+ registered users. Video-based learning; acquisitions globally. Financial difficulties in recent years illustrate challenges.³
Ruangguru: Indonesian EdTech platform; millions of users; video content and tutoring.
Yuanfudao / Zuoyebang (China): Chinese tutoring platforms with hundreds of millions of users. Affected by regulatory changes limiting tutoring.
African EdTech: Growing ecosystem including Eneza Education (Kenya), uLesson (Nigeria), Tuteria (Nigeria).
India EdTech ecosystem: Large number of startups beyond BYJU'S: Unacademy, Vedantu, Toppr, and others serving massive market.
AI Tutoring for Scale
Khan Academy (Khanmigo): AI tutor being adapted for global deployment; multiple language support; free access model.
OpenAI, Anthropic, Google: Foundation models that power global AI tutoring; API access enables new applications.
Startups targeting developing markets: Various companies building AI tutoring for specific regions and languages.
Connectivity Providers
Starlink: Satellite internet with global coverage potential; could reach areas without terrestrial infrastructure.
Mobile operators: Cell networks provide internet access; partnerships for educational data.
Government programs: National broadband initiatives in many countries; school connectivity programs.
The Promise of AI at Scale
Near-Zero Marginal Cost
The fundamental shift: Traditional education requires teachers proportional to students. AI tutoring requires development cost but near-zero marginal cost per student.
What this means: The same AI tutor that serves one student can serve one billion. Quality doesn't degrade with scale.
The implication: The economics of global education change fundamentally. What was impossible becomes possible.
Quality Without Dependency on Local Teachers
The current constraint: Educational quality depends on teacher quality; teacher quality is limited in many regions.
What AI changes: AI tutoring can provide consistent, high-quality explanation regardless of local teacher availability.
Not replacement but supplementation: AI tutoring augments whatever local teaching exists; provides floor of quality.
Language Accessibility
Current barrier: Quality educational content exists primarily in major languages (English, Chinese, Spanish).
AI capability: Large language models can translate and operate in hundreds of languages.
The possibility: AI tutors providing instruction in local languages worldwide—not just content translation but actual tutoring interaction.
Adaptivity at Any Level
Current problem: Curriculum assumes preparation levels that many students don't have. Students fall behind and never catch up.
AI advantage: AI tutoring can meet students where they are—assess current level and adapt instruction accordingly.
The possibility: Students at any level can receive appropriate instruction without stigma or structural barriers.
The Barriers to Global AI Education
Connectivity
The digital divide: ~33% of global population lacks internet access; rural and poor populations most affected.
Bandwidth requirements: AI tutoring requires reasonable connectivity; not available everywhere.
Cost of access: Even where available, internet may be unaffordable relative to income.
Solutions emerging:
- Satellite internet (Starlink) extending coverage
- Mobile networks expanding
- Zero-rating of educational content by carriers
- Offline-capable AI (limited capability but improving)
Devices
Smartphone penetration is high: ~80% of adults globally have mobile phones; ~70% have smartphones.
Computer access is limited: Many students lack computers for extended study.
Shared devices: In many households, devices are shared; limits individual access.
Quality variation: Older devices may not run current AI applications effectively.
Solutions emerging:
- Lower-cost devices optimized for learning
- School-based device programs
- Feature phone-compatible services
- Public access points (libraries, schools, community centers)
Language and Content
Major languages covered: AI works well in English, Chinese, Spanish, and other major languages.
Smaller languages underserved: Thousands of languages with limited AI capability.
Cultural context: Educational content often reflects Western/developed country contexts.
Local curriculum alignment: AI tutoring must align with local educational standards and expectations.
Solutions emerging:
- Multilingual model development
- Localization partnerships
- Community-generated content
- Local EdTech development
Affordability
Cost of AI: Current commercial AI APIs have costs that may be prohibitive at scale in low-income contexts.
Cost of devices and connectivity: Even if AI is free, access infrastructure costs.
Willingness to pay: Educational technology competes with other demands on limited budgets.
Solutions emerging:
- Subsidized access for education
- Open-source models reducing costs
- Public funding for educational AI
- Nonprofit and philanthropic support
Trust and Acceptance
Skepticism of technology: Some communities distrust technology-based education.
Teacher resistance: Teachers may see AI as threat and resist adoption.
Parental concerns: Safety, screen time, and quality concerns.
Institutional inertia: Educational systems slow to change.
Solutions emerging:
- Community engagement and co-design
- Teacher involvement and training
- Evidence building and sharing
- Gradual, supported adoption
New Inequalities
The AI Access Gap
Who benefits first: Students with devices, connectivity, and awareness of AI tools.
The pattern: Likely replicates existing digital divides. Urban over rural; wealthy over poor; educated parents over uneducated.
Risk: AI tutoring widens gaps in early adoption period before universal access achieved.
The AI Effectiveness Gap
Who uses AI well: Students with motivation, skills, and support to use AI effectively.
The pattern: Even with equal access, usage and benefit may vary. Students from advantaged backgrounds may extract more value.
Risk: Equal access could produce unequal outcomes, exacerbating rather than reducing inequality.
The Quality Gap
Premium vs. free: The best AI tutoring may be expensive; free versions may be inferior.
Human supplementation: Wealthy students may have AI plus human tutors and support; others have AI alone.
Risk: Two-tier system where AI tutoring is equalizing at basic level but elite education pulls further ahead.
The Skills Gap
What AI can't teach well: Creativity, social skills, physical skills, complex judgment.
The pattern: These skills may matter more in AI economy. Students with access to rich human education may develop them better.
Risk: AI equalizes cognitive skill development while elite institutions develop harder-to-automate capabilities.
The Credential Gap
Traditional credentials remain valued: Elite university degrees, personal networks, signals of privilege.
The pattern: AI education may not provide same credentials as traditional paths.
Risk: Even with equal learning, unequal credentials perpetuate inequality.
Policy Considerations
Ensuring Universal Access
Public provision: Governments could provide free AI tutoring to all students, like textbooks.
Infrastructure investment: Connectivity and devices as educational infrastructure, publicly funded.
Partnership models: Public-private partnerships to extend access.
International support: Development assistance focused on educational AI access.
Standards and requirements: Ensuring AI tutoring meets quality standards across all deployments.
Managing Quality
Quality assurance: Testing and certification of AI tutoring systems.
Accuracy standards: Requirements for educational AI to be accurate and aligned with curriculum.
Outcome monitoring: Tracking whether AI tutoring improves learning, not just engagement.
Continuous improvement: Mechanisms for identifying and correcting problems.
Protecting Vulnerable Populations
Child safety: Strong protections for AI interactions with children.
Privacy: Limits on collection and use of educational data.
Manipulation prevention: Preventing AI from influencing beliefs, values, or behaviors inappropriately.
Human oversight: Ensuring human supervision of AI education.
Supporting Transitions
Teacher roles: Policies for how teacher roles evolve as AI becomes prevalent.
Training and support: Preparing educators to work alongside AI.
Job impacts: Addressing potential displacement in education sector.
Cultural preservation: Ensuring AI doesn't homogenize education or undermine local culture.
International Coordination
Standards: Common standards for educational AI quality and safety.
Knowledge sharing: Sharing what works across countries and contexts.
Preventing regulatory arbitrage: Avoiding race to bottom on protections.
Supporting developing countries: Ensuring AI benefits reach all countries.
The Path Forward
Near-Term Likely (2026-2032)
AI tutoring spreads unevenly: Adoption accelerates in developed countries and urban developing country contexts. Rural and poor areas lag.
Major language coverage expands: AI tutoring available in dozens of languages; major global languages well served.
Quality varies: Best AI tutoring very effective; others less so. No consistent standards.
Digital divide persists: Connectivity and device access remain barriers for billions.
Pilot programs multiply: Governments and organizations test AI tutoring at scale in developing contexts.
Plausible (2032-2040)
Connectivity barriers largely overcome: Satellite and mobile networks provide global coverage. Cost falls.
AI tutoring becomes norm in most countries: School systems integrate AI; students expect AI support.
Learning gaps begin narrowing: For subjects AI teaches well, outcomes converge across contexts.
New skill gaps emerge: Gaps persist or widen in areas AI doesn't address well.
International standards develop: Common expectations for educational AI quality and safety.
Wild Trajectory (2040+)
Global educational floor raised dramatically: Every child with device access has access to high-quality instruction. Learning poverty plummets.
Traditional geographic inequality reduced: Place of birth matters less for basic educational access.
New elite education emerges: Human-intensive education becomes the privilege; AI tutoring is baseline.
Global labor market implications: Skill levels converge worldwide; labor competition intensifies.
Cultural implications: Global AI curriculum influences global culture; tension with local identity.
Case Studies
India: Scale and Ambition
Context: 1.4 billion people; hundreds of millions of students; dramatic variation in educational quality; strong EdTech ecosystem.
Current state: BYJU'S, Unacademy, and others serving millions. Government digital infrastructure (Aadhaar, UPI) could enable educational distribution.
Potential: India could be first country to deploy AI tutoring at truly massive scale—hundreds of millions of students.
Challenges: Quality control; device access; connectivity in rural areas; teacher integration; regulatory framework.
Implications: If India succeeds, model for other developing countries. If it fails, caution for others.
Sub-Saharan Africa: The Hardest Challenge
Context: Youngest population; highest proportion out of school; limited infrastructure; many languages; lowest connectivity.
Current state: Growing mobile penetration; early EdTech ecosystem; significant international investment.
Potential: If AI tutoring can work here, it can work anywhere. Leapfrog traditional educational infrastructure.
Challenges: Languages; connectivity; devices; electricity; conflict and instability; teacher capacity.
Implications: Tests whether AI education can truly reach the most underserved.
China: Central Planning at Scale
Context: World's largest educational system; strong state capacity; advanced AI development; centralized control.
Current state: Tutoring restrictions but educational technology permitted; AI development accelerating.
Potential: Government could deploy AI tutoring to all students as policy decision.
Challenges: Control over content; privacy and surveillance concerns; international access limitations.
Implications: Demonstrates centralized model for AI educational deployment.
The Deeper Questions
Who Decides What's Taught?
If AI tutors reach billions, who determines what they teach? The companies that build them? The governments that approve them? International bodies?
The concentration risk: A few AI providers could shape education for billions. Unprecedented influence over human development.
Cultural implications: Will AI education homogenize global culture? Reinforce dominant perspectives? Or preserve diversity?
The governance challenge: No framework exists for governing global AI education. Who has authority?
Is Equal Access Enough?
If everyone has access to AI tutoring, is the educational equity problem solved?
Clearly no: Access is necessary but not sufficient. Usage, effectiveness, human support, complementary resources all matter.
The risk: Declaring victory on access while inequality persists in outcomes.
The responsibility: Ensuring not just access but actual benefit for all students.
What's Lost in Scale?
Local knowledge, cultural specificity, human relationship, community connection—these may not scale with AI.
The trade-off: Global AI tutoring may provide more consistent quality while sacrificing local relevance.
The question: What's the right balance between scalable consistency and local specificity?
The answer: Likely varies by subject, age, and context. Not one-size-fits-all.
Does AI Education Serve Global or Local Interests?
Education shapes citizens and workers. Whose interests does AI education serve?
Colonial echoes: Education has historically been used to impose external cultures and create dependent labor forces.
Sovereignty concerns: Countries may resist AI education they don't control.
Labor market implications: AI-educated workforce in developing countries may compete with workers elsewhere.
The political economy: Who benefits from global AI education? Local populations? Global employers? AI providers?
Risks and Guardrails
Digital Colonialism Risk
Risk: AI education controlled by developed country companies, reflecting their perspectives and serving their interests.
Guardrails: Local control over educational content; open-source AI enabling local deployment; international governance frameworks; support for local AI development.
Surveillance and Control Risk
Risk: AI education systems collect massive data on children; governments or companies use for surveillance or manipulation.
Guardrails: Strong privacy protections; data localization requirements; limits on data use; transparency requirements.
Quality Failure Risk
Risk: AI tutoring spreads widely but doesn't actually improve learning—or actively harms through errors or manipulation.
Guardrails: Rigorous outcome evaluation; quality standards and certification; continuous monitoring; mechanisms for correction.
Equity Failure Risk
Risk: AI education exacerbates rather than reduces inequality—through access gaps, effectiveness gaps, or new skill gaps.
Guardrails: Universal access policies; targeted support for disadvantaged populations; monitoring for differential impact; equity-focused design.
Dependency Risk
Risk: Countries or communities become dependent on AI systems they don't control; loss of local educational capacity.
Guardrails: Building local capacity alongside AI deployment; maintaining human educational infrastructure; resilience planning.
Conclusion
The possibility before us is unprecedented: high-quality educational support available to every human, regardless of where they were born or what resources their family has. For the first time in history, the economics permit it. The technology enables it.
Whether it happens depends on choices society makes. Connectivity must extend to all. Devices must be affordable. AI must work in all languages. Quality must be assured. Access must be universal.
The alternative is equally clear: AI tutoring as another advantage for the already advantaged. Wealthy children with AI plus human tutors; poor children with nothing. The global educational divide widening rather than closing.
The stakes are immense. Education determines opportunity, determines what people can become and contribute. If AI tutoring reaches the half of humanity that has been underserved by traditional education, the potential unlocked is enormous. Human talent exists everywhere; educational opportunity does not. AI could change that.
But AI could also reinforce and extend existing inequalities. The same technology that enables democratization enables stratification. Which outcome emerges depends on policy, investment, and will.
The next decade will likely determine the trajectory. If AI tutoring scales equitably—reaching the schools without teachers, the villages without connectivity, the children without access—the result could be the most significant expansion of human capability in history. If it doesn't, humanity will have missed an opportunity that may not return.
The technology is arriving. The question is whether it gets deployed for all.
Endnotes — Chapter 31
- UNESCO data shows approximately 250 million children and youth not enrolled in primary or secondary education; concentrated in Sub-Saharan Africa, South Asia, and conflict-affected regions.
- World Bank "Learning Poverty" metric shows over 50% of children in low and middle-income countries cannot read and understand a simple story by age 10.
- BYJU'S valuation peaked at $22 billion in 2022; subsequently faced financial difficulties illustrating challenges in EdTech business models.
- Starlink global coverage expanding; educational partnerships in development; cost remains barrier for many potential users.
- Mobile phone penetration data from GSMA; smartphone penetration ~70% globally; feature phone usage remains significant in some regions.
- India EdTech market grew rapidly during COVID; government digital infrastructure (Aadhaar identification, UPI payments) provides foundation for digital service delivery.
- Sub-Saharan Africa has youngest population globally (median age ~18); highest rates of out-of-school children; fastest-growing internet penetration.
- Chinese tutoring restrictions (2021) limited commercial after-school tutoring but permitted educational technology for in-school use.
- Language coverage of major AI models typically includes ~100 languages with varying quality; thousands of languages have limited or no AI support.
- Global Partnership for Education funding supports education in 90+ countries; focus on lower-income countries and fragile states.