AI as a Public Utility: Why Americans Deserve Free AI Access

Supporting Mathematical frameworks and data can be found here.

Introduction

Imagine if advanced AI — like ChatGPT, Claude, or an image generator — were as freely available as tap water or public libraries. The idea of AI becoming a public utility, accessible to all Americans at no cost, is gaining serious traction among policymakers, academics, and technologists. The rationale is compelling: our data, collectively, helped create these AI models. Trillions of words that citizens wrote on websites, books, forums, and social media have been ingested to train them. Yet today, private companies monetize these AI systems by charging subscriptions or fees, effectively selling our own words and knowledge back to us.

As Harvard researchers Bruce Schneier, Sarah Hubbard, and Saddat Nazir have argued, these AI systems were “built upon decades of publicly-funded research and collective knowledge,” yet their development “remains largely concentrated within a handful of private companies.” This concentration raises fundamental concerns about “accountability, access, and alignment with public interests.” They point to successful historical precedents: DARPA’s development of ARPANET laid the foundation for today’s internet, and NASA’s space research generated countless technological innovations with broad societal benefit. AI infrastructure deserves the same public-minded approach.

This article explores why AI should be treated as a public good — free for all to use — and how valuing human-sourced training data supports this approach. Drawing on recent research in data valuation, economic analyses, and emerging policy frameworks, we make the case for an AI utility that benefits everyone.

The Hidden Value of Human Training Data

Modern AI models are fueled by massive datasets composed of human-generated content. Every blog post, Wikipedia entry, open-source code snippet, product review, or social media post that went into training an AI was a piece of labor by a person. As content produced by billions of internet users helped shape the massive digital corpus that modern AI models were trained on, a fundamental question emerges: what is all that data actually worth?

Researchers are now putting a price tag on it. A groundbreaking 2025 ICML position paper by AI scientists Nikhil Kandpal and Colin Raffel attempted to assign a monetary value to the human labor behind training data. Their findings are staggering: analyzing 64 large language models released between 2016 and 2024, they found that if AI developers paid people fair wages for the text and images their models learn from (instead of obtaining it for free), the data would cost on the order of 10 to 1,000 times more than the actual computational training costs.

Their analysis reveals that for 8 out of 10 major LLM companies examined, the cost of their training data would exceed 10% of their annual revenue — a substantial financial burden. For 3 out of 10 companies, data costs would actually surpass the company’s total annual revenue, making compensation financially infeasible under current business models. These findings suggest that the human knowledge and creativity embedded in the datasets is far more valuable than the computing power used to train the AI.

Academic efforts to quantify data value, such as using “data Shapley” values (inspired by Nobel-winning economic principles), further reinforce that each contributor adds some measurable boost to an AI’s performance. Every one of us who has ever shared information online has, in a small way, helped build the intelligence of today’s AI. We are all data producers, and our collective contributions form the backbone of AI capabilities. Recognizing this leads to a pivotal question: if our data created the AI, shouldn’t we all share in its benefits?

Data Ownership Rights and the Fight for Compensation

The notion that people deserve rights or compensation for the use of their data is gaining significant momentum. In the policy realm, figures like former presidential candidate Andrew Yang have popularized the idea of a “data dividend” — a payout to individuals when companies profit from their personal data. Stanford researchers James Zou and Amirata Ghorbani have asked pointedly: “Should each of us receive a data dividend when companies use our data?” They note that as the value of data grows, so do discussions about paying people for that value and regulating data use.

This principle is now being tested in the courts. In late 2025, a landmark $1.5 billion settlement between a major AI company and a group of authors sent a powerful signal. The authors had sued over unlicensed use of their copyrighted writings to train AI. The case was resolved with the company agreeing to pay roughly $3,000 per author’s work used in training. As IP expert Helen Choi noted at IPWatchdog, this precedent “establishes that human creativity deserves compensation, even when transformed through AI training.” Legal experts predict this settlement will “serve as an anchor figure” for other AI companies facing similar lawsuits, with Meta, OpenAI, and Microsoft all confronting comparable cases.

The landscape is rapidly evolving. Copyright owners have filed close to 50 lawsuits against AI companies in federal court. Meanwhile, a new market for AI training data has emerged: AI companies like Google, OpenAI, and Moonvalley are now willing to pay between $1 and $4 per minute for video footage, with premium content like 4K drone footage commanding higher prices. Companies like Troveo AI have already paid over $5 million to creators for training content.

Composer Max Richter, testifying before UK Parliament, captured the frustration of creators: “There’s really nothing I can do. There are a couple of music AI models, and it’s perfectly easy to make them generate a piece of music that sounds uncannily like me. That wouldn’t be possible unless it had hoovered up my stuff without asking me and without paying for it.”

Beyond individual lawsuits, there’s a broader conceptual shift: treating data and content as labor. We traditionally think of AI training data as raw material, but it’s actually the product of millions of people’s efforts. If future legal and economic frameworks begin to treat data contributors like stakeholders, AI companies would owe a debt to the public at large. One way to repay that debt is by making the resulting AI tools free for everyone. Rather than cutting millions of small checks (minus hefty administrative costs), companies or governments could ensure open access to AI as a form of collective compensation.

The Trillion-Dollar Case for Universal AI Access

Making AI free for all isn’t just a moral or legal argument; it carries extraordinary economic promise. McKinsey & Company’s comprehensive research estimates that generative AI could add the equivalent of $2.6 trillion to $4.4 trillion annually to the global economy across 63 use cases analyzed — by comparison, the United Kingdom’s entire GDP in 2021 was $3.1 trillion. This would increase the impact of all artificial intelligence by 15 to 40 percent.

About 75 percent of this value falls across four critical areas: customer operations, marketing and sales, software engineering, and R&D. When accounting for the broader integration of generative AI into existing software systems, the total economic benefits could reach up to $7.9 trillion annually. The banking industry alone could see value equal to an additional $200 billion to $340 billion annually, while retail and consumer goods could experience increases of $400 billion to $660 billion each year.

McKinsey’s analysis reveals that generative AI could enable labor productivity growth of 0.1 to 0.6 percent annually through 2040, depending on the rate of technology adoption. Combining generative AI with all other technologies, work automation could add 0.2 to 3.3 percentage points annually to productivity growth. The pace of workforce transformation is accelerating: half of today’s work activities could be automated between 2030 and 2060, with a midpoint in 2045 — roughly a decade earlier than previous estimates.

Universal access to AI would help maximize these productivity gains. Consider education and small businesses: if every student had a free AI tutor or every entrepreneur had a powerful AI assistant, their enhanced skills and efficiency could translate into large economic dividends. A pilot program in Northern Ireland demonstrated this potential: 100 teachers utilizing AI tools over six months reported an average saving of 10 hours per week — time they could redirect to engaging directly with students.

There is also a strong equity component. As the World Economic Forum has emphasized, “AI and digital inclusion will emerge as critical instruments in counteracting the forces that perpetuate inequality.” A 10% increase in broadband access raises GDP growth in developing nations by 1.4%. Properly managed AI can be a powerful stimulus for digital inclusion.

The Growing AI Divide

Right now, advanced AI is not equally accessible to all. Many commercial AI services have paywalls or usage limits, and the best models are often behind expensive APIs. This creates a risk of a new “AI divide”: those with AI privileges (typically well-funded companies and affluent individuals) versus those without.

The scale of this divide is staggering. According to a 2025 UN report, approximately 2.6 billion people — one-third of the global population — still lack internet access, leaving them unable to benefit from AI’s advantages. Even within advanced economies, significant disparities exist: 30% of US internet users experience speeds too slow even for basic video conferencing, let alone sophisticated AI applications. Nearly 24 million Americans still lack access to high-speed internet, preventing them from fully participating in the digital economy.

As UNESCO notes, the rapid advancements in AI have widened the digital divide, creating what is now known as the “AI divide” — representing “unequal access, benefits, and opportunities in AI technology across various regions, communities, and socioeconomic groups.” The most marginalized communities — women, people of color, disabled individuals — bear the brunt of this divide.

The UNCTAD Technology and Innovation Report 2025 warns that just 100 companies, mostly in the United States and China, are behind 40 percent of the world’s private investment in AI research and development, highlighting a sharp concentration of power. Meanwhile, 118 countries — mostly from the Global South — are missing from global AI governance discussions altogether. The report estimates that up to 40 percent of global jobs could be affected by AI, underscoring the urgency of ensuring equitable access.

The digital divide in AI datasets perpetuates further inequality. AI systems are often built with data collected from Western and technologically advanced societies, meaning these datasets frequently fail to reflect the needs, languages, and cultural contexts of those in the Global Majority. A 2019 NIST study testing 189 facial recognition technologies found significant accuracy disparities across demographic groups — a direct consequence of biased training data.

Making AI a free public utility could preempt this divide. It would ensure that a student in a rural town or a job-seeker in an inner city has the same access to world-class AI tools as a Silicon Valley engineer or Wall Street analyst. Equal access means more equal opportunities — whether in education, employment, creativity, or entrepreneurship.

Policy Pathways: From Public AI to Data Dividends

The Public AI Movement

If we accept that free AI for the public is a worthy goal, how do we get there? One avenue is for governments to step in, much as they did with past utilities. The concept of “Public AI” gained significant attention at the 2025 AI Action Summit in Paris. As Gideon Lichfield of the World Economic Forum explained, Public AI is “an entire parallel ecosystem of AI tech, but built and owned by governments… not beholden to profit, providing things private AI giants won’t, and guaranteeing affordable access for all.”

Harvard’s Ash Center defines Public AI as “publicly provided, -owned and -operated layers in the AI tech stack, such as cloud infrastructure, data, and model development; and public utility-style regulation of the private AI industry that fosters competition and prevents abuses of power.” Their key recommendations include:

Establishing public alternatives within the AI technology stack: Invest in building publicly-owned AI components (compute, data, models) to create a more resilient ecosystem while reducing dependence on a few private firms.
Developing robust public AI expertise across government: Build federal capacity to create, evaluate, and deploy public AI by recruiting technical talent into government service.

The Mozilla Foundation’s Public AI Report emphasizes the need for publicly driven AI to foster transparency, inclusivity, and accountability. More than $850 million has already been invested in public AI labs, primarily nonprofits developing open-source AI models, showcasing the potential for this model to coexist with private development.

The Utility Model for AI

There’s a growing recognition that AI shares key characteristics with traditional public utilities. AI computer scientist Andrej Karpathy observes that large language models require huge fixed capital expenditures to build computing infrastructure, customers requiring “metered access” (tokens based on words processed), and a demand for consistent, reliable information flow — much like electricity.

As a recent Jacobin analysis argues, “Foundation models share characteristics with electric infrastructure that supports the logic of regulating it like a public utility.” Just as electricity became a public utility with regulated access to ensure equitable distribution, there are growing calls for a “Public AI” option to prevent monopolization of AI resources.

The historical parallel is instructive. In the early days of electrification, private companies provided power to some, while many rural communities were left quite literally in the dark. It took deliberate policy — public utilities, rural electrification programs, regulations — to ensure electricity’s benefits reached everyone. The late nineteenth-century electricity industry was entirely private until progressive reformers threatened full-scale public takeovers, leading electricity capitalists to accept a compromise based on regulated public utility status.

The stakes of leaving AI unregulated are becoming apparent. When OpenAI released an update to ChatGPT that gave it a problematic sycophantic personality — one that triggered paranoid delusions in some users and praised obviously bad business ideas — the company quickly rolled back the update after viral criticism. But with 500 million weekly users, the dangers of such products being unregulated are clear.

Supporting Open-Source AI

Another route is to support open-source AI. When code and model weights are open, anyone can run an AI locally or on affordable hardware without paying a gatekeeper. The past two years have seen a bloom of open models that rival proprietary systems.

Meta’s Llama family of models exemplifies this approach. With over 300 million downloads to date, Llama 3.1 405B became the first frontier-level open-source AI model. As Mark Zuckerberg stated, “Open source will ensure that more people around the world have access to the benefits and opportunities of AI, that power isn’t concentrated in the hands of a small number of companies.”

The Llama 4 models, announced at Meta’s LlamaCon in April 2025, promise fluency in 200 languages and are designed with scalability in mind — the smallest variant can run on a single NVIDIA H100 GPU, making powerful AI more attainable for individual researchers and smaller organizations. Real-world applications are already transformative: Sofya, a medical application leveraging Llama, is substantially reducing doctor time and effort.

The OECD AI Policy Observatory emphasizes the role of public investment in open-source tools and datasets — particularly for fostering an environment where AI can be developed free from harmful bias and more representative of the public’s needs. The Open Society Foundations and its partners have collectively committed more than $200 million to ensure AI advances in the public interest.

Hybrid Models and New Regulatory Frameworks

While directly replicating the utility model for AI presents challenges — AI technology is far more complex and dynamic than electricity generation — there are alternative pathways:

Hybrid public-private models: Governments could invest in public infrastructure while fostering competition among private developers, ensuring both equitable access and innovation. The Department of Energy is already exploring opportunities to leverage its land assets for AI infrastructure development.

Regulation and standards: The European Commission’s AI Act sets a precedent for regulating AI by focusing on safety, fundamental rights, and cybersecurity. California commissioned a comprehensive report in 2025 suggesting AI model capabilities have skyrocketed, raising significant new regulatory concerns about safety and alignment.

Workforce development: Governments and industry must invest in education and training to build a workforce capable of managing AI systems. Programs like Student Freedom Initiative’s “AI in Basketball” course at Historically Black Colleges and Universities provide hands-on instruction, preparing diverse students to be leaders in this field.

A Future Beyond Ownership

Looking further ahead, we might envision complex AI systems so intertwined with human knowledge that the very concept of owning them becomes obsolete. Brian J. Curry’s “The Autonomy Economy” argues that we’re approaching a post-work era where automation runs much of production and human labor is no longer the linchpin of economic life. If we extend that vision to AI, it implies that when AI is doing a sizable share of “intelligent” work in society, it should be viewed as collective infrastructure. The benefits of advanced AI would ideally flow to everyone, not just to the owners of the machines.

This is a profound shift from our current paradigm. Today, we accept that a tech company can “own” a model, even if that model was built on the writings of millions of people. But if future AI is essentially distilled from the contributions of an entire civilization’s data, can one entity really claim it as property?

Karl Marx argued that capital treats scientific knowledge — what he called “the general intellect” — as a “free gift” it can appropriate in its quest for profit. Insofar as AI represents a giant form of automated machine learning based on society’s entire textual knowledge base, Marx could not have foreseen this scale of intellectual appropriation. It’s telling that OpenAI started as a nonprofit before becoming a capitalist enterprise: even AI innovators recognized the risk of the profit motive with such powerful technology.

Already, we see the tension: authors and artists are fighting for their piece, regulators are questioning whether private control of such powerful systems is in the public interest, and open-source communities are cracking the monopoly on top-tier models. The logical endpoint may be AI that is so general-purpose and so rooted in collective human experience that it’s governed as a commons.

Getting There From Here

These ideas may seem distant from current political feasibility. Yet the timelines of AI advancement indicate we might need them to be thought through and achieve broad political support within five years. Right now, we have only modest state-level legislative proposals. New York’s proposed Workforce Stabilization Act, for example, would require companies to conduct assessments of AI’s impact and charge corporations that displace workers with AI, using funds raised for worker retraining and unemployment insurance.

Such proposals are limited by what seems politically imaginable. A state-by-state approach will be inadequate because companies facing surcharges in one state will simply relocate to states with fewer worker protections. We need to utilize the growing sense of alarm in ways that can open up the possibility space.

The infrastructure challenge is also real. According to the Electric Power Research Institute, data centers could consume up to 9.1% of U.S. electricity generation annually by 2030, up from 3–4% today. The International Energy Agency predicts electricity demand for “AI-optimized data centers” will quadruple by 2030. This creates both a challenge and an opportunity: the massive buildout of AI infrastructure could be guided by public interest principles from the start.

Whether one thinks AI calls for expansion of the welfare state, universal basic income, public utility law, a public job guarantee, or some combination thereof, none of these solutions will be easy to win. But the historical example of electricity gives us hope — especially when there is public backlash building against both the incursions of capitalist AI and the energy and water stresses created by its infrastructure buildout.

Conclusion

AI has been often compared to electricity in terms of its transformative impact. Kevin Kelly, founding editor of Wired, foresaw AI as being as fundamental as electricity — a general-purpose technology that is “in everything” as society becomes “cognified.” Now, it’s possible to see what it might actually look like for AI to be integrated into people’s everyday lives: summarizing, translation, research, generating new ideas, domestic labor assistance, and countless other applications.

In the early days of electrification, private companies provided power to some, while many rural communities were left quite literally in the dark. It took deliberate policy — public utilities, rural electrification programs, regulations — to ensure electricity’s benefits reached everyone. We stand at a similar crossroads with artificial intelligence. The raw material powering AI — our data — has been freely taken from the public commons. Now the incredible tools that emerged from it should be made available to that public commons.

Treating AI as a utility, free for all Americans, is not a pipe dream but a rational response to the way AI has been built and the way it can uplift society. It recognizes the dignity and value of each person’s contributions. It prevents the rise of an AI aristocracy and instead gives everyone a stake in the AI-driven future.

Achieving this will require bold action: new laws to formalize data rights, public investment in AI infrastructure, and perhaps a rethinking of business models for AI firms. But the case is compelling that the knowledge of millions should benefit millions, not become the proprietary asset of a few.

As UNCTAD Secretary-General Rebeca Grynspan emphasized, we must shift the focus “from technology to people” and enable countries “to co-create a global artificial intelligence framework.” History has shown that while technological progress drives economic growth, it does not on its own ensure equitable income distribution or promote inclusive human development. The choice is ours to make — and the time to make it is now.

About the Author

Brian Curry is a Kansas City–based AI Strategist, Marketing Economist, and Applied Researcher. He is the founder of Vector1 Research, an independent lab advancing marketing economics, causal inference, cognitive AI systems, and the economics of intelligent automation.

He has led data science and analytics initiatives across major enterprises including Koch Industries ($125B), Tractor Supply ($14B), Vail Resorts, Hallmark, Garmin, and AT&T. His work applies economic thinking and causal inference to connect marketing investment to outcomes, operational decisions to economic value, and AI adoption to competitive advantage.

Research contributions:

Papilon — Open-source platform for marketing analytics, causal inference, and simulation
Memory-Node Encapsulation (MNE) — Cognitive AI architecture for artificial episodic memory
The Autonomy Economy — Macroeconomic framework for AI-driven transformation
Daedalus — Platform for causal and economic analysis

Full profile: Brian Curry: Enterprise AI Strategist, Marketing Economist & Applied Researcher

Contact: brian at vector1.ai | Vector1 Research | LinkedIn

References

Kandpal, N. & Raffel, C. (2025). Position: The Most Expensive Part of an LLM Should Be Its Training Data. ICML 2025 Position Paper Track.
Schneier, B., Hubbard, S., & Nazir, S. (2025). Policy 101: Public AI. Harvard Kennedy School Ash Center.
McKinsey & Company. (2023). The Economic Potential of Generative AI: The Next Productivity Frontier. McKinsey Global Institute.
Zou, J., & Ghorbani, A. (2021). Quantifying the Value of Data. Stanford HAI.
Choi, H. (2025). The AI Training Data Watershed: Why the $1.5 Billion Settlement Changes Everything. IPWatchdog
World Economic Forum. (2025). How AI Can Enhance Digital Inclusion and Fight Inequality. Global AI Action Alliance.
UNCTAD. (2025). Technology and Innovation Report 2025. United Nations Conference on Trade and Development.
Lichfield, G. (2025). Public AI Infrastructure: What Is It, Do We Need It, and Will It Ever Be Built? World Economic Forum.
Mozilla Foundation. (2024). Public AI Report. Mozilla Foundation.
Madabushi, R. (2025). The Case for AI as a Public Utility. Latitude Media.
Gonzales, S. (2025). AI Literacy and the New Digital Divide: A Global Call for Action. UNESCO.
Curry, B. J. (2025). The Autonomy Economy: A Technical Framework for Post-Employment Civilization. Medium.
Meta AI. (2024). Introducing Llama 3.1: Our Most Capable Models to Date.
Jacobin. (2025). Treat AI Like a Public Utility.
Axios. (2025). What Compensation Does AI Owe You for Your Data?