A Simple Thought Experiment: When Learning Becomes "Infringement"
Imagine a human being. You walk into a public library. You browse a few books. You skim chapters. You absorb ideas, patterns, and ways of thinking. You don't photocopy pages. You don't memorize paragraphs verbatim. You simply learn.
You go home and build something — a business idea, a piece of software, a research paper, a novel, a product design. Later, someone recognizes echoes of their work in what you created. They say: "My book influenced you. You learned from it. Therefore, you owe me a license."
Under U.S. copyright law, this claim fails immediately. Copyright does not protect influence, learning, or inspiration. It protects expression — not ideas, patterns, or knowledge absorbed through exposure. If this were not true, education itself would be illegal.
Continue Reading
This essay is available to subscribers. Join thousands of readers exploring faith, culture, and the life of the mind.
Why Humans Are Allowed to Learn — And Machines Suddenly Aren't?
The only meaningful distinction critics can point to is this: "But AI is a machine." That distinction is legally irrelevant. Copyright law has never hinged on who learns — only on whether protected expression was reproduced, whether the output substitutes for the original work, and whether the use is transformative.
A human reading a book does not "store" the book. An AI training on text does not "store" the book. In both cases the original expression is not retained, the output is new, and the process is transformative. Calling one lawful learning and the other infringement is not a legal principle — it's a policy preference.
The Absurd Consequences of Treating Learning as Copying
If influence alone were infringing, then:
- Every student would owe licenses to textbook authors
- Every scientist would need permission to cite prior research before thinking
- Every engineer would need a license for concepts absorbed from documentation
- Every novelist would be legally exposed to anyone whose style influenced them
Copyright would collapse under its own weight. The law avoids this outcome by drawing a firm line: ideas, facts, styles, and knowledge are free to learn from. AI training operates squarely in this protected space.
Why This Analogy Matters Legally
Courts have already endorsed this logic.
If reading to learn is legal for humans, then learning to model language must be legal for machines — or the law has abandoned coherence.
The Real Question Courts Must Answer
The issue is not whether AI was influenced by copyrighted works. Of course it was. So is every educated person alive. The real question is:
Does the system reproduce protected expression in a way that harms the market for the original work? If the answer is no, then training cannot be infringement — no matter how uncomfortable that makes legacy business models.
Why Getting This Wrong Breaks the Legal System
If courts reject this analogy, they will be forced to police influence instead of copying, quantify "how much learning is too much," and decide which ideas are "licensed thoughts." That path leads to an unworkable, incoherent copyright regime.
The law was never meant to govern cognition — human or artificial. And if it tries, it will fail.