The Great Data Heist
How AI and Social Media Companies Extracted Trillions in Value from Human Data — and Left the Humans with Nothing
March 2026 | An Investigative Analysis
The combined market capitalization of the five largest technology companies — Apple, Microsoft, Alphabet, Amazon, and Meta — now exceeds $14 trillion, built substantially on the collection, analysis, and monetization of data generated by billions of ordinary people. In 2025 alone, Google and Meta together earned over $460 billion in advertising revenue, almost entirely derived from user-generated behavioral data for which users received no direct compensation. Meanwhile, a new generation of artificial intelligence companies has scraped the public internet — harvesting text, images, code, and creative works produced by hundreds of millions of individuals — to train models now valued in the hundreds of billions of dollars. This article examines the ethical, moral, and legal implications of this data extraction economy, quantifies the value transferred from individuals to shareholders, and assesses the growing legal and regulatory reckoning now underway.
1. The Bargain You Never Agreed To
Every time you scroll through Instagram, search on Google, watch a YouTube video, or post on Facebook, you are creating economic value. Your clicks, your pauses, your likes, your location data, your browsing history, your social connections, your photographs, your messages — all of it is recorded, analyzed, packaged, and sold. The companies that facilitate these interactions have grown into the most valuable enterprises in human history. Yet the billions of people whose daily lives generate this wealth receive nothing in return beyond access to platforms that are, in economic terms, the most lopsided bargains since the colonial era.
This is not hyperbole. Research from the University of Pennsylvania found that the median Facebook user values their personal data at approximately $750 per year. A 2025 analysis by Proton, drawing on public financial disclosures, estimated that an American user generates at least $700 annually for Google and Meta combined. Yet these users receive precisely zero dollars in compensation.
The fundamental exchange — free services in return for unlimited data extraction — was established in the early 2000s and has gone largely unquestioned for two decades. But the scale of the resulting wealth transfer has now grown so enormous, and the asymmetry so stark, that scholars, regulators, and the public are beginning to ask a question that should have been asked long ago: is this arrangement ethical, or is it simply theft dressed in the language of innovation?
2. Quantifying the Extraction: The Numbers Behind the Data Economy
The Advertising Machine. The digital advertising industry is the primary mechanism through which human data is converted into corporate revenue. In the fourth quarter of 2025 alone, Google generated $82 billion in advertising revenue while Meta earned $58 billion — a combined $140 billion in a single quarter, derived almost entirely from targeting capabilities built on user data.
For the full year 2025, Meta reported total revenue of $201 billion, of which approximately 97% came from advertising. The company’s 3.58 billion daily active users each generated an average of roughly $54–56 in annual advertising revenue globally — a figure that rises to approximately $235 for American users.
Alphabet’s advertising revenue for 2025 is estimated at approximately $265 billion across Search, YouTube, and its broader ad network. Together, Google and Meta generated over $460 billion in advertising revenue in a single year. Global digital advertising revenue is projected to have exceeded $1 trillion in 2025.
The AI Gold Rush. Layered on top of the advertising economy is a new wave of value extraction: artificial intelligence. OpenAI’s annualized revenue surged from $200 million in early 2023 to $13 billion by August 2025. Anthropic’s revenue grew from $87 million in early 2024 to $7 billion by late 2025 — an approximately 80-fold increase in under two years. OpenAI is now privately valued at approximately $300 billion, which would rank it among the top 15 companies globally if it were publicly traded. Anthropic is valued at more than $60 billion. These valuations rest substantially on models trained on data scraped from the open internet — text written by journalists, code written by developers, art created by illustrators, conversations held by ordinary people — none of whom were asked for permission, and virtually none of whom received compensation.
A Back-of-the-Envelope Calculation. If we conservatively estimate that 2 billion internet users in developed economies each generate $700 per year in value for Google and Meta alone, the annual value extracted from these users amounts to approximately $1.4 trillion. Over the past decade, the cumulative value extracted from user data by the major platforms likely exceeds $8–10 trillion — wealth that flowed overwhelmingly to shareholders, executives, and a comparatively small number of employees. The five largest technology companies returned over $1 trillion in stock buybacks and dividends in 2025 alone.
Key figures at a glance: Meta’s full-year 2025 advertising revenue was approximately $195 billion. Alphabet’s was approximately $265 billion. Meta averaged 3.58 billion daily active users. Each American Meta user generated roughly $235 per year in ad revenue. Proton Research estimated the combined annual value of a US user’s data to Google and Meta at $700. OpenAI reached $13 billion in annualized revenue, Anthropic reached $7 billion, and Big Tech collectively spent over $405 billion on AI infrastructure in 2025. OpenAI’s private valuation reached approximately $300 billion.
3. The Moral Case: Data as Stolen Labor
The ethical framework for evaluating this extraction is straightforward, even if the implications are uncomfortable. When a factory employs workers to produce goods, the workers are compensated for their labor. When a mining company extracts natural resources, it typically pays royalties to the landowner or the state. But when technology companies extract behavioral data from billions of people — data that is the essential raw material of their business — the producers of that data receive nothing.
This asymmetry has drawn comparisons to historical resource extraction. The Economist observed in 2017 that data had become more valuable than oil. But unlike oil, data is extracted not from the ground but from human beings — from their habits, preferences, relationships, movements, health conditions, and innermost thoughts. It is, in a meaningful sense, a form of unpaid labor.
Jaron Lanier, the computer scientist and author, has been among the most articulate critics of this arrangement. In his book “Who Owns the Future?” he argued that the digital economy systematically concentrates wealth by destroying the economic value of information created by ordinary people. When a photographer’s images train an AI model, when a writer’s articles are scraped to build a language model, when a musician’s compositions are ingested by an AI music generator — value is being transferred from creator to corporation without consent or compensation.
Andrew Yang, during his 2020 presidential campaign, proposed the concept of a “data dividend” — a payment to citizens from the companies that profit from their data. Research from the University of Pennsylvania’s Data Dividend Project found that users affiliated with data advocacy groups valued their Facebook data at a median of $1,000 per year, significantly higher than the general population’s median of $750. Crucially, when users were given information about how much their data was actually worth, 98% of those who revised their valuations revised them upward.
4. The AI Training Data Crisis: Scraping Without Consent
The ethical concerns surrounding data extraction have intensified dramatically with the rise of generative AI. Training a large language model or image generator requires vast quantities of data, and AI companies have obtained this data primarily by scraping the public internet — often in direct violation of website terms of service, copyright law, and basic norms of consent.
An investigation revealed that companies including Meta, Microsoft, and Nvidia extracted over 15.8 million videos from more than 2 million YouTube channels without obtaining consent from content creators, using this material to train generative AI video models. As one affected creator asked: “Should I keep making things in the hope of connecting with people, or just stop altogether?”
The International Association of Privacy Professionals (IAPP) has documented how a class action lawsuit against OpenAI alleged that the company used “stolen private information, including personally identifiable information, from hundreds of millions of internet users, including children of all ages, without their informed knowledge or consent.” The legal claims invoked the Electronic Communications Privacy Act, the Computer Fraud and Abuse Act, the Illinois Biometric Information Privacy Act, and multiple state consumer protection statutes.
The scale of these lawsuits is staggering. The New York Times sued OpenAI and Microsoft for using millions of articles without permission. A class of authors secured a $1.5 billion settlement from Anthropic over allegations of unauthorized use of copyrighted books. Clearview AI agreed to a $50 million settlement for scraping billions of facial images. Google reached a $1.6 billion settlement with Singular Computing over AI-related patent infringement. Meta was hit with a $1.4 billion settlement by the State of Texas over biometric and location data collection.
The lawsuit scoreboard so far: Anthropic settled with authors for $1.5 billion over copyright claims (settlement challenged). Meta settled with Texas for $1.4 billion over biometric and location data violations in May 2025. Google settled with Texas for $1.4 billion over data privacy violations the same month. Google settled with Singular Computing for $1.6 billion over AI patent infringement. Clearview AI settled for $50 million over facial image scraping in March 2025. The New York Times’ copyright case against OpenAI remains pending, as do lawsuits against Apple over book training data and against LinkedIn over private messages used for AI training.
5. The Legal Reckoning: From Gray Area to Courtroom
The legal landscape around data extraction is shifting rapidly. For years, AI companies operated in what many described as a legal gray area — technically, much of the data scraped from the internet was publicly accessible, and the application of copyright law to machine learning training was untested. That era is ending.
A landmark moment came in early 2025 when a federal judge in Thomson Reuters v. Ross Intelligence ruled that AI training on copyrighted legal content was not protected by fair use, finding that the use was commercial, not transformative, and harmful to the potential market for AI training data. This ruling sent a clear signal: publicly available does not mean freely usable.
The European Union’s AI Act, which began taking effect in phases, represents the most comprehensive regulatory framework to date. It introduces a risk-based classification system for AI, strengthens protections around biometric data and copyrighted content, and requires AI developers to maintain detailed records of their training data. Under the GDPR, which the AI Act supplements, companies must have a lawful basis to collect and process personal data — even when that data comes from publicly accessible sources.
In the United States, the Generative AI Copyright Disclosure Act of 2024 would require AI companies to disclose the datasets used to train their models. The No AI FRAUD Act aims to prevent unauthorized AI impersonation of individuals. Multiple states are pursuing aggressive enforcement: Texas alone secured settlements exceeding $2.8 billion from Meta and Google in 2025 over data privacy violations.
Perhaps most consequentially, the Khan v. Figma case filed in late 2025 reframed AI data risk not as a copyright issue but as a consent, contract, and trade secret problem. The lawsuit alleged that Figma used customer design files to train AI tools after years of assuring users their content would not be repurposed — a shift accomplished through buried policy changes rather than affirmative consent. Legal analysts have noted that this framing may prove more consequential than any single copyright ruling.
6. The Concentration of Wealth: Who Benefits?
The wealth generated by the data economy is concentrated to a degree that would have been unimaginable a generation ago. The five most valuable technology companies — Apple, Microsoft, Alphabet, Amazon, and Meta — account for more than 70% of the total market value of the top 20 global companies, up from 65% the prior year.
In 2025, Nvidia’s market capitalization had grown more than 800% since January 2023, driven by demand for AI chips. Microsoft, Amazon, Alphabet, Apple, and Meta each surpassed $2 trillion in market value. US stock buyback authorizations set a new record in 2025, rising above $1 trillion for the first time. Within this total, Alphabet repurchased $62.8 billion of its own shares, Meta repurchased $48.2 billion, and Apple deployed $100 billion — capital flowing directly to shareholders and executives rather than to the users whose data generated the underlying revenue.
Meanwhile, Big Tech capital expenditure on AI infrastructure exceeded $405 billion in 2025, with projections approaching $600 billion for 2026. Goldman Sachs projects total hyperscaler capital expenditure from 2025 through 2027 will reach $1.15 trillion. This spending is financed increasingly through debt — Meta and Oracle alone issued $75 billion in bonds in September and October 2025 — yet it is justified by the expectation that user data, now flowing into AI models, will generate even greater returns.
The stark reality is this: the five hyperscaler companies are expected to generate $577 billion in operating cash flow in 2026, up from $378 billion in 2023. This extraordinary wealth creation rests on a foundation of data produced by billions of ordinary people who have never been offered a share of the proceeds.
7. The Path Forward: Data Dividends, Regulation, and Digital Rights
Several frameworks have been proposed to address the fundamental inequity of the data economy.
Data Dividends. California Governor Gavin Newsom has proposed data “dividends” that would compensate consumers for the value created from their digital footprints. The concept builds on Andrew Yang’s 2020 proposal and is supported by growing academic research demonstrating that users systematically undervalue their data. If Meta’s approximately 250 million US users each received their proportional share of the company’s US advertising revenue ($100–104 billion), each user would receive approximately $400–416 per year. For a household of four, that represents $1,600 or more annually — a meaningful sum for many American families.
Consent and Transparency Requirements. The EU AI Act and pending US legislation point toward a future in which companies must disclose their training data sources and obtain meaningful consent before using personal data for AI training. The IAPP has documented that under the GDPR, explicit consent requirements apply to personal data scraped from the web, regardless of whether that data is publicly accessible. The challenge lies in enforcement.
Collective Bargaining for Data Rights. Some scholars have proposed treating data producers as a form of labor collective, with negotiating power over how their data is used and compensated. This model draws on the precedent of intellectual property licensing and could be facilitated by blockchain-based data marketplaces that give individuals control over their own information.
Strengthened Copyright Protection. The Thomson Reuters ruling and the wave of ongoing litigation suggest that courts are increasingly willing to apply existing copyright frameworks to AI training data. If this trend continues, AI companies will be forced to license the data they use or face crippling legal liability — a development that could fundamentally reshape the economics of the industry.
8. Conclusion: The Reckoning Has Begun
The technology industry’s extraction of value from human data represents one of the largest wealth transfers in economic history. Over the past two decades, a handful of companies have built multi-trillion-dollar empires on data produced by billions of people who received nothing in return. The rise of generative AI has intensified this extraction, as companies scrape the creative and intellectual output of hundreds of millions of individuals to build models that now compete directly with the people whose work made them possible.
The moral case against this arrangement is clear: people have a right to share in the value created from their data, their creativity, and their labor. The legal case is strengthening, with billions of dollars in settlements already secured and landmark court rulings establishing that publicly accessible does not mean freely exploitable. The economic case is compelling: redistributing even a fraction of the hundreds of billions generated annually from user data would provide meaningful income to millions of families.
The companies that built the data economy will argue that they created the platforms, bore the risk, and invested the capital. This is true. But it is also true that their platforms would be worthless without the billions of people who populate them with their lives. The question is no longer whether these people deserve compensation. The question is how much, and how soon.
Brian James Curry is an Enterprise AI Strategist, Data and Marketing Scientist, and Applied Researcher. He is the founder of Vector1 Research, an independent research lab focused on marketing economics, causal systems, cognitive AI architecture, and the economic implications of intelligent systems. He also founded KC AI Lab, a data science community and incubator for applied machine learning education.
References and Sources
1. Collis, Moser, & Whinston (2022). “Information Frictions and Heterogeneity in Valuations of Personal Data,” Penn Carey Law — Economics of Digital Services; Proton (2025). “What’s Your Data Really Worth?”
2. The Keyword (Feb. 2026). “Google and Meta Lead Q4 with $82B and $58B in Ad Revenue.”
3. MEXC News (Mar. 2026). “Meta’s $201 Billion Revenue Year”; Datapods (2025). “What Your Data Is Actually Worth.”
4. Alphabet FY2025 financial estimates via The Keyword, Visual Capitalist, and MEXC News.
5. GroupM / MarTech (Dec. 2024). “Global Ad Revenue to Top $1 Trillion.”
6. Visual Capitalist (Nov. 2025). “Charted: The Soaring Revenues of AI Companies (2023–2025)”; Epoch AI (Apr. 2025).
7. Bain & Company (2025). “Will AI Disrupt Tech’s Most Valuable Companies?” Technology Report 2025.
8. S&P Global Market Intelligence (Jan. 2025). “US Market Dividends in 2025”; MUFG Americas (Jan. 2026).
9. Collis, Moser, & Whinston (2022). Penn Carey Law EODS study; Data Dividend Project.
10. SEOBot AI (Sept. 2025). “Investigation Reveals Unauthorized Data Scraping from YouTube for AI Training.”
11. IAPP. “Training AI on Personal Data Scraped from the Web.”
12. The Lyon Firm (2024–2025); Traverse Legal (May 2025). “Recent Lawsuits Against AI Companies.”
13. Copyright Alliance (Jan. 2026). “AI Copyright Lawsuit Developments in 2025.”
14. USC IP & Technology Law Society (Feb. 2025). “AI, Copyright, and the Law.”
15. Traverse Legal (May 2025). Texas settlements against Meta and Google.
16. Rain Intelligence (Jan. 2026). “When AI Customer Data Training Becomes a Consent Problem.”
17. Bain & Company (2025). Technology Report 2025.
18. S&P Dow Jones Indices (2025). US buyback data.
19. I/O Fund (Nov. 2025). “Big Tech’s $405B Bet”; BofA Research; IEEE ComSoc Technology Blog (2025).
