The rise of Artificial Intelligence (AI) has triggered a crisis unlike anything we’ve seen before. Unlike the assembly lines that replaced physical labor, AI threatens to separate economic value from human effort entirely.
The loudest voices in the room — Silicon Valley venture capitalists like Marc Andreessen — push a philosophy called “Techno-Optimism.” They argue that if we let technology rip without regulation, it will inevitably create abundance and lift everyone up.
This view is economically naive and historically illiterate.
Without a counter-force, the AI transition will act as a massive “wealth pump,” transferring trillions of dollars from the people who work (labor) to the people who own the machines (capital). But there is hope. Buried in the technical research on how these models work is a secret vulnerability: AI desperately needs high-quality, human-generated data to survive.
This biological dependency gives us leverage. But that leverage is temporary.
Divided, we fall. If we allow any safe haven for exploitation, we ensure suffering for the many. We need a “Union of Humans” — a collective bargaining entity not just to protect jobs, but to redefine work and assert that if machines run on our data, we should be paid for it.
The Line Must Go Up, But for Whom?
The Productivity Bargain
For decades, the deal was simple: as technology made workers more efficient, wages went up, and goods got cheaper. This was the “productivity bargain.” The “Techno-Optimist Manifesto” reiterates this, claiming that because human wants are infinite, technology will always generate new forms of employment, and markets will naturally distribute the gains of productivity. But over the last 40 years, that deal has been broken. While productivity skyrocketed, wages stayed flat. The gains went to the top. AI will only make us more of what we already are, and faster, and what we are now is a system that does not work for the many.
AI is poised to accelerate this trend. The International Monetary Fund (IMF) warns that in advanced economies, 60% of jobs could be impacted by AI. When software can write code, analyze laws, or handle accounting, the demand for human experts drops, even as the work gets done faster.
It’s important to remember: Whatever story about strawberries or extra fingers you have pocketed to give yourself permission to disengage, this is the worst it will ever be. We are still in the Wright Brothers era of AI, and Anthropic already sees us at a precipice. Pretending there’s no monster in the closet is not a strategy.
Inequality as a Feature, Not a Bug: The Wealth Gap
This creates a “wealth pump” where the gains from efficiency accrue almost exclusively to the owners of the AI models and the hardware infrastructure. As noted by researchers at the NBER, falling labor shares are directly associated with rising income inequality. Most of us sell our labor to live; the wealthy own capital that makes them more wealthy. In an AI world, capital becomes everything, and labor becomes optional, even annoying to deal with.
In an AI-dominated economy, where “intelligence” is a capital asset owned by a few corporations, the vast majority of humans, lacking ownership of these assets but essential to their success, face a future of permanent economic marginalization. The IMF warns that in advanced economies, about 60 percent of jobs may be impacted by AI, with half of those facing reduced labor demand and lower wages.
A major driver of this inequality is something economists call “monopsony power.” You know what a monopoly is (one seller). A monopsony is the opposite: one buyer.
In the digital world, Big Tech is the only buyer of your data. You create the raw material — your posts, your clicks, your art — that trains their algorithms. In exchange, you get “free” services like search or social media. But while your data creates trillions in value for them, you get paid zero. Because they control the market, you have no power to negotiate a better price. You receive the equivalent a Visa gift card for your digital person-hood.
Frictions such as non-compete agreements and the specialized nature of tech platforms lock users and workers into ecosystems where they have no bargaining power. If a freelance writer or artist wants to participate in the modern economy, they must use platforms that scrape their work to train the very AI systems designed to replace them. This is the definition of a coercive market failure. The individual worker, or “data producer,” has zero leverage against a trillion-dollar entity. Only a collective entity — a union — can correct this imbalance.
The Crisis of Care
As AI aggressively automates makes cognitive work (like writing or coding) cheap, you might think “human” jobs — like nursing, teaching, or caregiving — would become more valuable. Economists call this “Baumol’s Cost Disease”: services that require human time should get more expensive as other things get cheaper.
But in a ruthless market, the opposite might happen. As writers and coders lose their jobs, they may flood into care work, driving wages down for everyone. We risk a future where the most essential human tasks — caring for the sick, teaching children — are the lowest paid, while the automated sector generates untaxed billions.
Research confirms that AI cannot replicate the empathetic components of these roles. For instance, while AI can grade papers, it cannot “inspire a struggling student” or manage a lively classroom discussion with the effectiveness of a human. A collective bargaining agreement for humanity must address the valuation of labor that cannot and should not be automated. It must demand that the AI dividend (more on this later) be used to subsidize and elevate the wages of the care economy, recognizing that “efficiency” is not the only metric of civilization.
The Idea of Data as Labor
“Data Dignity” and Liberal Radicalism: We are the Serfs
Thinkers like Jaron Lanier and Glen Weyl argue for a radical shift: we need to treat Data as Labor. Currently, the internet treats data as “exhaust” — waste that corporations can scoop up for free. This is a feudal arrangement. We are serfs working the land (the internet) for the lords (Big Tech). We live here, but we own nothing.
Lanier and Weyl argue that this “free data” model is a market failure. It treats data as capital owned by the firm rather than labor provided by the user. Every time a user solves a CAPTCHA, tags a photo, writes a Reddit comment, or uploads code to GitHub, they are performing work that trains an AI model. Without this continuous stream of human input, the models would stagnate.
The “Data as Labor” movement proposes that if we organize into “data unions” or “data coalitions,” we can collectively negotiate the terms of access to our digital lives. This shifts the framework from “privacy” (hiding our data) to “property” (selling or licensing our data on our terms). This approach also addresses the fear of automation: if data is treated as labor, then as AI uses more data, the demand for “data labor” (and thus the income going to humans) should theoretically rise, rather than falling.
Your every habit across every site is tracked. A digital profile of you is built and used to steer you towards purchases, sites, donations, and ideas. That digital profile is sold off in ways none of us fully understand. Everyone profits except you, the lab rat led through the maze to the cheese.
The Mechanics of Extraction
The need for ownership is clear. For years, Reddit users built a massive library of human conversation for free. Then, in 2024, Reddit sold access to that data to Google for millions to train AI. The users — the people who actually created the value — got nothing.
The users who created the value received nothing. The community, which acted as the de facto labor force, was bypassed entirely in the transaction between the platform owner and the AI developer. Users pointed out the ethical breach: “Remember when we all agreed to give our data to be sold for AI training? Me neither.”
Similarly, visual artists have been the “canaries in the coal mine” for AI displacement. Generative models like Midjourney and Stable Diffusion were trained on billions of copyrighted images scraped from the web without consent. This has led to the emergence of proto-unions and resistance tools. Tools like “Glaze” and “Nightshade,” developed by researchers at the University of Chicago, allow artists to subtly alter their images so they look normal to humans but “poison” AI models trying to copy their style. These are the digital equivalents of throwing a wrench in the gears — a way to force a negotiation.
Ghost Work in the Global South
We also can’t forget the invisible humans already powering AI: the data labelers and RLHF (Reinforcement Learning from Human Feedback) workers. This work is often outsourced to the Global South (Kenya, Philippines, India) where workers review toxic content, label images, and correct AI outputs for pennies on the dollar. This is the same work Facebook moderators once said gave them PTSD.
UNI Global Union and other labor organizations have identified this as a critical area for international solidarity. A “Union of Humans” cannot be a club for Western creatives; it must be a global entity that prevents the race to the bottom by standardizing wages for data labeling and RLHF work worldwide. If the Global South is left out, Big Tech will simply move the extraction offshore, undermining the bargaining power of workers in the Global North.
The Fragility of the Machine is the Leverage
Scaling Laws and the “Data Wall”
The fatalism surrounding AI act like AI progress is inevitable, but it’s not. Technical research into Scaling Laws reveals a critical vulnerability. Large Language Models (LLMs) improve effectively only when fed exponentially larger amounts of high-quality data. They need massive amounts of new, high-quality human text.
Estimates suggest that the stock of high-quality public human text (books, articles, high-quality code) is finite and will likely be exhausted by AI training runs between 2026 and 2032. The “low-hanging fruit” of the internet — Wikipedia, Reddit, Common Crawl — has already been harvested. They cannot invent new culture, new slang, or new history. They need us to keep creating it.
This “data scarcity” is the primary leverage point for a Human Union. The AI companies need us. They need our new data. If a global union were to organize a “data strike” — where millions of people simultaneously withheld new content, privatized their feeds, and used tools like Nightshade — the progress of AI would not just slow; it would hit a hard ceiling.
Can’t they just use AI to write data for other AIs? No. When AI models train on AI-generated content, they suffer from Model Collapse. The models start to “inbreed,” losing nuance and eventually spewing gibberish. This proves that humanity is the ground truth. We are the reference standard that keeps the machines sane.
Reinforcement Learning from Human Feedback (RLHF)
The current generation of useful AI relies heavily on Reinforcement Learning from Human Feedback (RLHF) to align the model with human values and utility. This process involves armies of human workers rating AI responses, correcting errors, and writing “gold standard” examples.
Without RLHF, an LLM is an unruly autocomplete engine; with RLHF, it becomes a helpful assistant. This dependence on human feedback loops is likely permanent. As models get more capable, the feedback required to train them becomes more complex, requiring higher-skilled human annotators (PhDs, lawyers, coders).
A global union would organize these RLHF workers, who are the literal teachers of the AI. If the teachers walk out, the student stops learning. The unionization of the “human-in-the-loop” pipeline is the most direct path to seizing control of AI development.
There is a non-zero chance this changes one day. Some of the most dedicated, brightest minds are dedicating themselves to solving AGI because they believe it will solve humanity’s problems. This highlights the urgency to act collectively now. I believe these bright minds would agree compromising with the masses to avoid disruptions is a drop in the bucket compared to the riches they promise.
A History of Resistance and Organization
Reclaiming the Luddite Legacy
The Luddites of the 19th century weren’t anti-technology; they were pro-worker. They smashed weaving machines because factory owners were using them to bypass labor standards and produce junk products. Their resistance eventually forced the British government to pass labor laws.
Today’s “Anti-AI” movement shares this DNA. It is not about banning calculators; it is about opposing the unauthorized expropriation of human creativity. The lesson from history is that disruption is often the only way to force capital to the negotiating table. You can’t pick the pocket of a starving artist on an IOU as you pour billions into data center expansion.
The WGA Strike: A Blueprint for Victory
The most successful modern example of collective bargaining against AI is the 2023 Writers Guild of America (WGA) strike. The writers faced an existential threat: studios wanted to use LLMs to generate scripts and hire writers only to “polish” them at lower rates, effectively turning writers into editors of machine output.
The WGA held the line for 148 days, leveraging their control over the production bottleneck. They won a historic contract that explicitly regulated AI usage.
This victory proved that collective action works against rampant industry. Through solidarity, the writers forced the studios to subordinate the technology to human needs. The challenge now is to scale the “WGA model” from a few thousand writers to the entire human population.
Considerations for the “Union of Humans”
Universal Basic Dividend
We must reject the “Universal Basic Income” (UBI) narrative pushed by tech oligarchs like Sam Altman (Worldcoin), which frames the payout as charity for the obsolete. It positions humans as passive recipients of aid.
Instead, we demand a Universal Basic Dividend (UBD) derived from the ownership of the data. The model is the Alaska Permanent Fund. In Alaska, the oil belongs to the people, and the oil companies pay a royalty into a fund that distributes a dividend to every resident. The fund uses a “Percent of Market Value” (POMV) draw to ensure sustainability.
Data is the new oil. The AI companies are the extractors. The Union demands that a percentage of the compute revenue be paid into a Global Data Fund. This is not charity; it is a return on the capital — our collective human intelligence.
Overcoming Fragmentation: The North-South Alliance
The greatest threat to this union is scabbing — specifically, the exploitation of workers in the Global South to train models for pennies, as with RLHF. The “Union of Humans” must be aggressively internationalist. It must ensure that RLHF workers in Kenya are paid parity wages with data licensors in New York. If the Global South is left out, Big Tech will simply move the extraction offshore.
Existing international solidarity networks provide a framework for this alliance. The Union must subsidize the organization of workers in developing nations, recognizing that a weak link anywhere is a vulnerability everywhere. History shows that trade unions face obstacles in international solidarity due to competition for work, but the unique nature of AI — where data quality is global — aligns the incentives of workers across borders more than in manufacturing.
The Cultural Shift: Revaluing Each Other
Finally, the Union must lead a cultural shift. We must reject the Techno-Optimist assertion that speed and efficiency are the only virtues. We must champion friction, nuance, and human connection. We must learn our lessons from the social media era, where irresponsible and unregulated tech didn’t provide a more connected world, as was promised. We must argue that a teacher is not just a content delivery system replaceable by an iPad, but a mentor. A doctor is not just a diagnostic engine, but a healer. An artist is an expression of the world, not its statistical average.
By asserting the value of the human touch, we protect the jobs that Baumol’s Cost Disease threatens. We make “Human Made” a status symbol, a moral choice, and an economic necessity.
Conclusion: A Binary Choice
We stand at a fork in the road of history. One path leads to Techno-Feudalism: A world where a handful of trillionaire warlords own the AI gods, and the rest of humanity subsists on their generosity, reduced to whatever is most cost-efficient to keep us alive and pacified by algorithms that feed us synthetic entertainment while our agency withers.
The other path leads to Human Solidarity: A world where the immense wealth generated by AGI is captured by the collective bargaining power of the species that created it. A world where technology serves us, because we own the fuel that powers it.
The leverage exists. The science proves they need us. The WGA strike proves we can win.
The time for individual adaptation is over. No individual, team, company, industry, country or continent is protected on their own. The time for collective bargaining has begun. We must form the Union.
Christian is prioritizing being a solution for AI alignment and caring for his neighbors. He has a personal website, which you can find here. He used AI in the making of this post. He is not fearful of the technology, only the systems that will exploit it. If you agree, please reach out on LinkedIn.