A Response to the AI Training Debate from the Humanities
Many of my friends and colleagues in the humanities are convinced that training Large Language Models on books, music, and art constitutes theft of intellectual property. I understand the anger. I share the underlying concern: that powerful institutions are extracting value from human creativity without compensating the humans who produced it. That’s a real injustice, and anyone who dismisses it isn’t paying attention.
But I want to suggest that the injustice runs deeper than the technology, and that the solutions most commonly proposed will make things worse for the very people they’re meant to protect.
The Anxiety Is Real
Let me start with what my friends and colleagues get right. Under our current economic system, creators are squeezed from every direction. Musicians earn fractions of pennies per stream. Adjunct professors with doctoral degrees teach for near-poverty wages. Freelance writers compete in a race to the bottom. When a technology company builds a billion-dollar product using the collective output of human culture, and the humans who produced that culture receive nothing, the outrage is justified.
Where I part company with my friends is not on the diagnosis of the disease, but on the proposed cure.
The “Theft” Metaphor Doesn’t Hold
The word theft implies that something has been taken away from its owner. If I steal your bicycle, you no longer have a bicycle. But information doesn’t work this way. If an LLM processes your novel during training, you still have your novel. Every copy remains intact. No reader has been deprived of access to your work.
This isn’t a trivial distinction. It reflects a fundamental property of information that economists call non-rivalry: one person’s use of an idea, a pattern, or a digital file does not diminish anyone else’s ability to use it. Physical goods are rivalrous; for example, if I eat your bread, you go hungry. But ideas, once expressed, can be shared infinitely without loss.
What happens during AI training is not reading in a human sense. It is computational analysis of the commons, the identification of statistical patterns across millions of texts. Humanists already have a name for this kind of work: distant reading, the term Franco Moretti coined for the computational study of literature at scale. Corpus linguists, digital humanities scholars, and literary analysts routinely process vast archives of text to identify patterns no individual reader could perceive. We have never considered this a violation of authors’ rights.
There is, of course, a crucial difference. When a digital humanities scholar performs distant reading, the output is scholarship, a meta-text about the corpus. When an AI company trains an LLM, the output is a generative system that can produce new texts, potentially competing with the original authors in the marketplace. That difference matters, and I don’t want to minimize it. But it is a difference in what is done with the patterns, not in whether extracting patterns from public culture is legitimate. The proper response to harmful commercial applications is to regulate the applications, not to criminalize the act of learning from the commons itself.
Who Actually Benefits from Stronger IP?
When creators call for stricter intellectual property enforcement against AI, I understand the impulse. But let’s consider who has historically benefited most from aggressive copyright regimes.
Not independent artists. Not poets or adjunct scholars. The primary beneficiaries of expanded IP law have consistently been large corporations such as Disney, major record labels, pharmaceutical companies, and academic publishers. These entities use copyright and patent law to hoard culture, extract rents from access to knowledge, and sue smaller creators who build on existing traditions.
If we establish the legal precedent that AI can only be trained on explicitly licensed data, we don’t stop AI development. We guarantee that only the wealthiest corporations (those that can afford to buy out the world’s copyright holders) will be permitted to build it. We hand the future of human knowledge to a cartel of licensors.
This is the historical pattern of enclosure: the privatization of what was once common. In early modern England, common lands that sustained peasant communities were fenced off for the profit of landowners. Today, the digital commons faces enclosure from two directions simultaneously: IP maximalists want to fence off the inputs (restricting what can be read, analyzed, or learned from); and tech monopolies want to fence off the outputs (ingesting the open web into proprietary, closed-source models and charging subscription fees to access the synthesized result). Both are forms of enclosure. Both concentrate power. And neither serves the independent creator.
Artificial Scarcity as a System of Control
To understand what’s really at stake, we need to name the economic structure that makes this debate so agonizing.
Capitalism is a system designed to allocate scarce resources through markets. When a resource is genuinely scarce (e.g., land, food, housing) markets serve a real (if often unjust) allocative function. But when a resource is infinitely replicable (e.g., a song, an algorithm, a medical formula) its natural market price trends toward zero.
This is a problem for capitalism. If the price of information is zero, no one can extract profit from it. So the system imposes artificial scarcity: paywalls, DRM, patents, licensing fees, and aggressive copyright enforcement. These mechanisms force inherently abundant resources to behave like scarce ones, so they can be bought and sold.
The result is a world in which a child in rural Nova Scotia and a child in rural Senegal may both lack access to the same textbook, not because the textbook is scarce, but because someone has decided it must be sold rather than shared.
When we demand stronger IP protections, we are (however unintentionally) reinforcing this system. We are insisting that ideas must be treated as property, that knowledge must be enclosed, that culture must be commodified. And the people who suffer most from that commodification are not tech executives. They are the poor, the marginalized, and the geographically isolated, precisely the people the humanities claim to care about.
“But Creators Need to Eat”
This is the strongest objection, and it deserves a serious answer rather than a hand-wave.
Yes. Creators need to eat. Right now, under our current system, intellectual property is one of the few tools available to individual creators trying to survive in a market that systematically undervalues their work. I am not suggesting that artists unilaterally surrender their copyrights while tech billionaires continue to accumulate wealth. Think of it as harm reduction: IP is a flawed tool, a tool of enclosure that overwhelmingly benefits corporations, but it is currently the only shield most creators have. The point is not to ask creators to drop that shield today. The point is to locate where the problem is actually situated, so we can build the structures that make the shield unnecessary, in order to eventually set it down together.
What I am suggesting is that the problem is not the free flow of information. The problem is an economic system that conditions human survival on the ability to monetize one’s output. The solution is not to re-enclose the commons. The solution is to change the conditions under which people create, and to do so with enough urgency that “wait for a better system” does not become an excuse for inaction.
Several approaches deserve serious consideration:
Universal Basic Income. If creators’ basic material needs (e.g., housing, food, healthcare) are met unconditionally, the desperate need to monetize every creative act evaporates. Artists can create for the sake of meaning, community, and human flourishing rather than market survival. If you think this is utopian fantasy, consider Canada’s own Mincome experiment in Dauphin, Manitoba in the 1970s (one of the most rigorously studied guaranteed income pilots in North America) demonstrated improvements in health, education, and community participation. More recent programs from Finland to Stockton, California, have shown that guaranteed income increases creative output and entrepreneurial risk-taking.
Taxing the monopolies. If AI companies are generating enormous wealth by synthesizing human culture, that wealth should be taxed and redistributed, not by locking down information, but by ensuring the public shares in the returns. A data dividend, an automation tax, or a corporate windfall tax directed into public arts funding would compensate creators without restricting the flow of knowledge.
Public funding for culture. Canada already does this better than most countries through the Canada Council for the Arts, SSHRC, and provincial arts councils. The principle is sound: treat cultural production as public infrastructure, funded collectively, available to all. We don’t need a new principle so much as a dramatic expansion of existing ones.
Open access as liberation, not theft. The Open Access movement in academic publishing demonstrates that freely available research accelerates discovery, benefits researchers in lower-income countries, and does not destroy scholarly careers, especially when paired with institutional support. Open Educational Resources have brought university-level instruction to millions who could never afford tuition. Rather than threats to human dignity, they are expansions of it.
These modern debates over access and enclosure are, in fact, echoes of much older moral questions.
A Voice from the Tradition
I write as a scholar of biblical literature, and I would be remiss not to note that the traditions I study have something to say about enclosure and abundance.
The Torah’s vision of jubilee (Leviticus 25) imagines a periodic restoration of economic relations, a systematic undoing of accumulated inequality. Land returns to its original holders. Debts are cancelled. The logic is that the earth ultimately belongs to God, not to any human proprietor: כִּי־לִי הָאָרֶץ “the land is mine” (Lev 25:23).
The prophetic tradition consistently frames the hoarding of resources as an offense against divine justice. Amos denounces those who “trample the head of the poor into the dust of the earth” (Amos 2:7). Isaiah warns against the monopolizers who “join house to house” and “add field to field, until there is no more room” (Isa 5:8), an ancient critique of enclosure that resonates uncomfortably with the modern enclosure of the digital commons.
The wisdom tradition, meanwhile, presents knowledge itself as a public good. Proverbs personifies Wisdom as calling out “in the street” and “at the busiest corner” (Prov 1:20–21) — not behind a paywall or inside a licensing agreement. The assumption is that wisdom, by its nature, seeks to be shared.
These are not proof-texts for a particular economic policy. But they do suggest that the instinct to enclose knowledge, i.e., to restrict access for the sake of private accumulation, stands in tension with some of the deepest currents of the Western moral tradition.
The Harder Question
I want to be honest about what I’m not claiming.
I’m not claiming that AI companies are benevolent. Maybe some are; I don’t know. I do know that the concentration of AI development in a handful of corporations is a genuine threat to democratic life, and it requires aggressive antitrust action, regulation, and public investment in open alternatives.
I’m not claiming that the transition to a post-scarcity information economy will be painless. It won’t. Real people are losing real income right now, and “wait for UBI” is not an adequate response to someone who can’t make rent this month.
And I’m not claiming that all forms of artificial scarcity are illegitimate. For example, privacy is a form of artificial scarcity applied to personal data, and it’s essential. Verifiable credentials require controlled access to maintain trust. The infrastructure that transmits digital goods (I’m talking about servers, cables, electricity) is genuinely scarce and requires management.
What I am claiming is this: when we respond to the disruption of AI by demanding that information be locked down more tightly, we are fighting the wrong battle. We are using the master’s tools to shore up the master’s house. The enclosure of knowledge has never served the vulnerable, and it will not start doing so now.
Conclusion
Information wants to be free. Stewart Brand famously added that information also wants to be expensive, which is true because the human labour required to create it is costly. Both halves of that tension are real, and I am not pretending otherwise. The question is not whether we can make information free; we can’t eliminate the cost of creation. The question is whether we will build economic structures that bear that cost collectively, allowing human beings to thrive in a world of informational abundance, or whether we will cling to artificial scarcity and condemn creators to an endless, losing war against the nature of digital goods.
Our goal should not be to put chains on information. It should be to free the humans who create it.