A Position Paper and Roadmap for Universal AI Access in the United States and Around the World
Executive Summary
Human civilization has repeatedly been transformed when a resource that was once scarce, expensive, or available only to the privileged became broadly accessible.
Electricity changed how humanity worked and lived. Clean water and sanitation transformed public health. Telecommunications connected societies. The Internet democratized access to information.
I believe artificial intelligence represents the next—and potentially most consequential—step in this progression.
AI should no longer be viewed merely as another technology product or commercial software service. It is rapidly becoming an essential intellectual infrastructure for modern society.
Just as electricity provides access to energy, AI provides access to intelligence, knowledge, analysis, problem-solving, and increasingly personalized assistance.
In the coming decades, the ability to use AI will affect a person’s ability to learn, obtain healthcare information, find employment, create businesses, manage finances, communicate, conduct research, interact with government, and participate fully in society.
For these reasons, I believe governments should recognize universal access to a basic level of artificial intelligence as an essential public service of the twenty-first century.
This does not mean governments should nationalize AI companies or replace private innovation.
Quite the opposite.
The successful model should combine:
Universal access + private-sector competition + public infrastructure + consumer protection + responsible regulation + continuous innovation.
The objective should be straightforward:
Every person, regardless of income, geography, age, education, or social status, should have access to a safe and capable basic AI service.
The United States should lead the world in establishing this model.
I. THE SOUTH KOREAN PRECEDENT
South Korea has recently taken an important step toward this concept through its national “AI for All” initiative.
The program is intended to provide citizens access to general-purpose AI services without usage limits, supported initially through government-provided advanced computing resources and subsequently through national funding.
The longer-term objective is particularly significant:
one AI agent for every citizen.
South Korea is also integrating AI with government services so that AI becomes not merely a chatbot but an interface through which citizens can interact with public institutions.
This represents an important philosophical shift.
AI is beginning to move from being considered simply a commercial technology product toward becoming part of a country’s essential digital infrastructure.
The United States and other nations should take this concept substantially further.
II. WHY AI SHOULD BE TREATED LIKE ESSENTIAL INFRASTRUCTURE
Consider what electricity provides.
Electricity does not tell us what appliances we must buy. Government does not manufacture every refrigerator, computer, television, or electric automobile.
Instead, society created an infrastructure that makes electricity reliably available, established technical standards, regulated safety and distribution, protected consumers, and allowed thousands of private companies to innovate on top of that infrastructure.
AI can follow a similar model.
The government does not need to build the dominant AI model.
It needs to ensure that the AI infrastructure exists, remains competitive, is safe and resilient, and provides every citizen with reasonable access.
There is an important distinction between electricity and AI.
Electricity is a physical commodity.
AI is an evolving intellectual capability.
Therefore, AI should not simply be regulated using the traditional public-utility model.
Instead, we should create a new category:
Essential Digital Intelligence Infrastructure
This framework would recognize AI as infrastructure fundamental to economic and social participation while maintaining a competitive marketplace for AI providers.
III. THE CASE FOR UNIVERSAL AI ACCESS
The strongest argument for universal AI access is not technological. It is economic and humanitarian.
Imagine two students of equal intelligence.
One has unlimited access to advanced AI tutoring, research, translation, writing assistance, mathematics, science instruction, career counseling, and personalized education.
The other has none.
Within several years, the difference between their opportunities could become enormous.
The same will increasingly apply to workers.
An employee using AI effectively may accomplish in hours what another employee requires days to accomplish.
Small businesses equipped with AI will have capabilities previously available only to large corporations—market research, accounting assistance, legal information, software development, advertising, customer service, and data analysis.
AI can similarly expand access to health information, government programs and scientific knowledge.
The danger is therefore not simply that AI might eliminate certain jobs.
A much greater societal danger may be that people with AI become dramatically more capable than people without AI.
We already speak of the digital divide.
Without intervention, we could create something considerably more consequential:
The Intelligence Divide.
Preventing that divide should become a major objective of public policy.
IV. A NATIONAL AI UTILITY FRAMEWORK FOR THE UNITED STATES
I propose that the United States establish a National Universal AI Access Framework based on seven pillars.
PillAR 1 — Universal Basic AI Access
Every American should have access to a baseline level of capable AI service.
The service should include reasonable access to:
- general knowledge and research assistance;
- education and tutoring;
- language translation;
- employment and career assistance;
- government information;
- small-business assistance;
- basic financial education;
- health and wellness information, subject to appropriate safeguards;
- accessibility assistance for elderly and disabled citizens; and
- personal productivity tools.
Basic access should either be free or available at nominal cost.
Commercial providers would remain free to offer substantially more powerful premium services.
This is comparable to having public libraries while simultaneously maintaining bookstores, publishers, universities, and commercial information services.
Universal access expands markets rather than eliminating them.
V. DO NOT CREATE A GOVERNMENT AI MONOPOLY
The United States should avoid creating a single government-controlled AI model.
A government monopoly could become technologically stagnant and could create legitimate concerns involving political influence, censorship, privacy, surveillance, and control over information.
Instead, government should establish standards and allow multiple qualified providers to compete.
Americans might therefore select among several certified AI providers.
The federal government could negotiate a basic national AI-access tier with participating providers, similar in principle to other public-private infrastructure arrangements.
Users wishing to purchase advanced capabilities could continue subscribing directly to commercial providers.
Competition would remain central to the system.
VI. CREATE AN AI UNIVERSAL SERVICE FUND
America has previously used universal-service concepts to expand communications infrastructure.
A comparable AI Universal Service Fund could ensure that economic circumstances do not determine access to artificial intelligence.
Funding could come from a combination of:
- Congressional appropriations;
- public-private partnerships;
- competitive government procurement;
- contributions associated with very large commercial AI infrastructure;
- spectrum or technology revenues where appropriate; and
- state and local participation.
The objective should not be to subsidize unlimited premium AI consumption.
It should establish a minimum national AI capability available to everyone.
Children from low-income families should have access to the same basic AI educational capabilities as children from wealthy families.
A small business in rural America should have access to AI productivity tools comparable to businesses in Silicon Valley.
VII. BUILD THE PHYSICAL INFRASTRUCTURE
AI may appear to exist in the cloud, but ultimately it depends upon enormous physical infrastructure:
electricity + semiconductors + data centers + telecommunications + cloud infrastructure + computing capacity.
The United States therefore needs a coordinated national infrastructure strategy.
AI policy cannot be separated from energy policy.
The country will need considerably greater electricity generation, transmission capacity, data-center infrastructure, advanced semiconductor manufacturing and high-speed communications.
Federal and state permitting processes should be modernized so strategically important AI and energy infrastructure can be built rapidly while maintaining appropriate environmental and community protections.
The government should establish National AI Infrastructure Zones in regions capable of supporting large concentrations of computing infrastructure.
These could combine:
data centers → energy generation → energy storage → advanced cooling → fiber connectivity → semiconductor infrastructure → research universities → workforce development.
The resulting infrastructure could become the twenty-first-century equivalent of America’s interstate highway system.
VIII. CREATE A NATIONAL COMPUTE GRID
The next step should be even more ambitious.
America should develop a National AI Compute Grid.
Today, advanced computing power is concentrated primarily among a relatively small number of large technology companies.
Universities, researchers, startups, physicians, nonprofit organizations, schools, and small businesses frequently cannot obtain comparable computing resources.
A National Compute Grid could provide shared access to qualified users through government facilities, national laboratories, universities, and participating private cloud providers.
Users could obtain compute credits based upon purpose and need.
Special allocations could support:
scientific research, medicine, education, national security, climate science, engineering, entrepreneurship and socially valuable research.
This would democratize access to the raw computational power underlying artificial intelligence.
IX. AI SHOULD BECOME PART OF GOVERNMENT ITSELF
Universal AI should not simply answer questions about government.
It should eventually help citizens interact with government.
Every citizen should have access to an AI Public Services Assistant capable of explaining and navigating appropriate government services.
Instead of searching through hundreds of government websites, citizens could ask:
“What benefits am I eligible for?”
“How do I apply for Social Security?”
“What tax forms do I need?”
“How do I start a small business?”
“What permits do I need?”
“What veterans’ benefits are available?”
“What educational assistance can my child receive?”
The AI assistant could guide citizens through the appropriate government process.
Government would become considerably more accessible.
AI could therefore produce not merely technological efficiency but democratic efficiency.
X. EDUCATION MUST BECOME AI-NATIVE
Providing AI without teaching people how to use it would be comparable to distributing books without teaching people to read.
AI literacy should therefore become a basic educational competency.
Schools should teach students:
how to communicate with AI;
how to evaluate AI-generated information;
how to verify sources;
how to recognize hallucinations and misinformation;
how to protect personal information;
how to use AI creatively;
and how to combine AI capability with independent human reasoning.
The objective should not be to make children dependent upon AI.
It should be to teach them how to think better with AI.
Every teacher should similarly have access to AI training.
Every community college should provide AI-literacy programs.
Libraries should become community AI learning centers.
Workers displaced or transformed by AI should receive publicly supported AI retraining.
XI. THE AI BILL OF RIGHTS AND RESPONSIBILITIES
Universal AI access must be accompanied by strong protections.
Citizens should have clearly established rights concerning:
Privacy — personal information should receive appropriate protection.
Transparency — users should know when they are interacting with AI.
Security — systems serving the public must meet cybersecurity requirements.
Choice — citizens should not be forced to use one AI provider.
Portability — users should be able to move appropriate personal AI information between providers.
Human review — consequential governmental decisions should have appropriate human appeal or review mechanisms.
Protection of children — age-appropriate safeguards should apply.
Accountability — organizations deploying high-risk AI should have identifiable responsibility for their systems.
Non-discrimination — AI used in consequential decisions should be evaluated for unlawful discrimination.
Due process — government should not allow opaque algorithms to become the final unchallengeable authority over citizens.
The existing NIST AI Risk Management Framework provides a useful technical foundation upon which these protections can be expanded.
XII. REGULATE APPLICATIONS ACCORDING TO RISK
We should avoid regulating every AI application identically.
An AI system recommending restaurants does not require the same regulatory framework as AI helping diagnose cancer.
I recommend three broad categories.
Tier I — General AI
Writing, research, entertainment, productivity, translation and routine consumer applications.
Regulation should be light.
Tier II — Consequential AI
Employment, education, insurance, lending and important government services.
Additional transparency, documentation and auditing should apply.
Tier III — High-Risk AI
Healthcare diagnosis, autonomous transportation, critical infrastructure, weapons, aviation, nuclear systems and other applications capable of causing significant physical or societal harm.
These systems should be subject to much stronger testing, validation, cybersecurity and human oversight.
The governing principle should be:
Regulate the risk created by the application, not AI simply because it is AI.
XIII. WHAT THE U.S. GOVERNMENT SHOULD CHANGE
America’s current AI strategy properly recognizes innovation, infrastructure and international leadership as national priorities.
But I believe another pillar should be added:
UNIVERSAL AI ACCESS.
Congress and the Executive Branch should jointly establish a national objective that every American have access to basic AI capability.
A National AI Infrastructure and Universal Access Act could establish:
- Universal Basic AI Access.
- A National AI Compute Grid.
- An AI Universal Service Fund.
- AI literacy programs.
- National interoperability standards.
- Privacy and portability standards.
- Risk-based AI regulation.
- Federal AI procurement standards.
- AI infrastructure permitting reforms.
- Public AI research resources.
- AI access programs for small businesses.
- AI accessibility programs for elderly, disabled, rural and economically disadvantaged populations.
Responsibility should be coordinated across the White House, Commerce Department, NIST, Department of Energy, Department of Education, Department of Labor, FCC, NSF, national laboratories and appropriate sector regulators.
However, America should avoid creating another enormous bureaucracy.
A small National AI Infrastructure Council could coordinate existing agencies rather than duplicate them.
XIV. FIVE-YEAR U.S. ROADMAP
YEAR 1 — DECLARE AND DESIGN
Congress recognizes basic AI access as an essential component of national digital infrastructure.
Establish the National AI Infrastructure Council.
Define minimum Universal Basic AI standards.
Develop national safety, privacy, interoperability and portability standards.
Launch pilot programs in selected schools, libraries, healthcare systems, rural communities and government agencies.
YEAR 2 — BUILD
Launch the AI Universal Service Fund.
Begin National Compute Grid deployment.
Accelerate data-center, transmission and power-generation permitting.
Expand broadband into remaining underserved areas.
Train teachers, federal employees and workforce-development personnel.
YEAR 3 — UNIVERSAL ACCESS
Provide every American access to a basic AI assistant.
Integrate federal public-service information.
Provide AI access through schools, universities, libraries and community centers.
Launch dedicated AI programs for seniors, veterans, disabled citizens and low-income communities.
YEAR 4 — INTEGRATION
Enable AI assistants to interact securely with participating government systems.
Expand health, education, employment and small-business applications.
Introduce secure personal-data portability.
Create an interoperable personal AI ecosystem.
YEAR 5 — AI-NATIVE AMERICA
Every citizen can have a personal AI assistant.
Every student has an AI tutor.
Every teacher has an AI teaching assistant.
Every small business has access to AI productivity tools.
Every researcher can apply for advanced compute resources.
Every citizen can navigate government through AI.
And every American can participate meaningfully in the AI economy.
XV. ROADMAP FOR OTHER DEVELOPED COUNTRIES
Developed economies should adopt similar principles while adapting implementation to their governmental structures.
Countries with national healthcare systems could integrate AI into patient education and navigation.
Countries with centralized education systems could deploy universal AI tutoring rapidly.
The European Union could create cross-border AI infrastructure and compute capacity while maintaining its emphasis on fundamental rights.
Japan, Canada, Australia, Singapore, the United Kingdom and other technologically advanced nations could establish interoperable national AI-access systems.
Ultimately, allied democracies should develop common standards allowing trusted AI systems and digital identities to operate internationally.
XVI. DEVELOPING COUNTRIES REQUIRE A DIFFERENT MODEL
Developing countries should not attempt simply to reproduce America’s enormous AI infrastructure.
Their priority should be access rather than ownership of frontier models.
The World Bank has identified four foundations necessary for widespread AI adoption:
Connectivity
Compute
Context
Competency
This framework is particularly appropriate for developing nations.
Reliable electricity and affordable Internet connectivity must come first.
Countries should then obtain affordable shared compute through regional cloud infrastructure rather than attempting independently to construct enormous frontier-AI systems.
AI systems must support local languages and local cultural, educational, agricultural and healthcare contexts.
Finally, populations need the skills necessary to use AI productively.
Developing countries should particularly emphasize lightweight AI capable of operating through inexpensive smartphones and other widely available devices.
XVII. CREATE REGIONAL AI UTILITIES FOR DEVELOPING COUNTRIES
International institutions could help establish regional AI infrastructure.
The World Bank and regional development banks could finance Regional AI Compute Centers serving multiple countries.
For example:
East Africa could share regional compute infrastructure.
West Africa could establish another network.
South Asia could create interconnected facilities.
Latin America could develop regional AI clouds.
Small island nations could access internationally hosted sovereign AI capacity.
This would be considerably more economical than expecting every country to build frontier-scale computing infrastructure independently.
International development assistance should increasingly include:
electricity + broadband + cloud access + AI compute + local-language models + AI education.
These are becoming fundamental development infrastructures.
XVIII. AI COULD REDUCE GLOBAL INEQUALITY
AI presents an extraordinary paradox.
Poorly distributed, it could dramatically increase inequality.
Universally distributed, it could become one of the greatest equalizing technologies ever created.
A child in a remote village could have access to an AI tutor capable of explaining mathematics and science.
A farmer could receive agricultural guidance.
A community healthcare worker could access medical knowledge.
A small entrepreneur could receive business assistance.
A student could learn another language.
A physician could consult global medical knowledge.
A government employee could deliver public services more efficiently.
For the first time in history, advanced intellectual assistance could potentially become available to billions of people at very low marginal cost.
That opportunity should not be underestimated.
XIX. A GLOBAL AI ACCESS INITIATIVE
I propose that the United States work with other nations and international organizations to establish a:
GLOBAL AI ACCESS INITIATIVE
Its long-term objective should be:
Every human being should have reasonable access to artificial intelligence by 2035.
This initiative could mobilize governments, technology companies, universities, development banks and philanthropic organizations.
The program would focus particularly on countries lacking electricity, broadband, computing capacity and digital education.
America should lead this initiative.
Doing so would simultaneously advance humanitarian development, American technological leadership and democratic values.
XX. WHO PAYS?
Universal access does not mean unlimited taxpayer funding.
The infrastructure should follow a mixed economic model.
Government should finance public goods and address market failures.
Private industry should compete to provide technology and advanced services.
Large commercial users should pay market rates.
Basic consumer access could be publicly supported.
Premium services could remain subscription-based.
Universities and researchers could receive compute grants.
Developing nations could receive international development financing.
This creates a sustainable ecosystem rather than a permanent government entitlement to unlimited computing resources.
XXI. THE ECONOMIC RETURN
The correct question is not:
“How much will universal AI cost?”
The more important question is:
“How much productivity will society lose if millions of people cannot participate in the AI economy?”
If AI meaningfully increases productivity across education, healthcare, science, manufacturing, government, professional services and small business, investments in universal access could generate economic returns far exceeding their costs.
We did not build electrical grids merely because people wanted electric lights.
We built them because electricity became the foundation upon which an enormous modern economy could develop.
AI should be viewed similarly.
XXII. THE LARGER VISION
Human beings have historically been constrained by access to knowledge and expertise.
A person could only know what he or she had learned, read, experienced, or could obtain from another person.
Artificial intelligence fundamentally changes that limitation.
Increasingly, every human being can have access to an intellectual assistant capable of drawing upon enormous bodies of human knowledge.
That development could become as consequential to human intellectual capacity as electricity was to human physical productivity.
I therefore believe society should stop asking whether AI is simply another technology industry.
We should begin asking a more important question:
What infrastructure does civilization require when intelligence itself becomes a universally accessible resource?
South Korea has taken an important early step toward answering that question.
The United States should go further.
America should combine its extraordinary private-sector innovation with a national commitment to universal AI availability.
Developed countries should establish comparable systems.
Developing countries should be assisted in building the electricity, connectivity, compute, local knowledge and skills necessary to participate.
The objective should not be government control of artificial intelligence.
It should be democratization of artificial intelligence.
Just as modern society came to recognize that electricity, clean water, communications and education should not be privileges available only to the wealthy, the coming AI society should recognize that access to intelligence-enhancing technology cannot become a privilege reserved for those who can afford it.
AI should become infrastructure for humanity.
And our generation has an opportunity—and, I believe, a responsibility—to make certain that this extraordinary technology improves the lives not merely of millions, but ultimately of every human being.
— Dr. Mohan Ananda
Founder, DRAI Health
Scientist • Entrepreneur • Policy Innovator