The Hedge · May 20, 2026 · Brutal Honesty Over Hype Since 2008
Copyright law was built on two assumptions: that expression is scarce, and that copying is detectable. AI has demolished both — and the legal system has no clean answer for what comes next.
A model can now ingest a novel, internalize its structure, voice, and ideas, then produce something functionally equivalent without reproducing a single protected sentence. That’s not a loophole. That’s a structural failure in the entire framework.
The cases on the docket
The litigation now working through the system — the NYT’s suit against OpenAI, the Authors Guild actions, Getty Images versus Stability AI — isn’t really about copying in the old sense. The central questions are whether training on copyrighted work constitutes infringement, and whether AI output competes with the original in the marketplace.
That second prong is where fair use gets complicated fast. The four-factor test has always weighted “market harm” heavily. If AI output replaces demand for the original, the “transformative use” defense takes serious damage — no matter how technically different the output is.
“Fair use was designed for humans doing creative work. An AI processing 100 billion tokens of human writing to produce commercial output doesn’t fit that mold — and courts know it.”
Where the doctrine breaks down
Fair use was designed for humans with expressive intent: a critic quoting a passage, a scholar analyzing a text, a parody riffing on an original. The doctrine assumes a person on the other end. That assumption is gone.
Courts will have to either stretch the doctrine until it’s unrecognizable, or acknowledge it simply doesn’t apply the same way. Neither path is clean.
The rewrite problem
Here’s the sharper issue: if AI can take any copyrighted work and produce a “better” version — cleaner prose, updated facts, same ideas — what exactly does copyright protect anymore?
Under current law: expression, not ideas. You can’t copyright a plot structure, a chord progression concept, or a journalistic angle. You can only copyright the specific words, notes, or images. A perfect paraphrase has always been legal. AI just industrializes it at a scale that makes the distinction feel hollow — because it is hollow, at that scale.
How this likely resolves
Four scenarios are on the table:
- Licensing regimes emerge. Like ASCAP/BMI for music — a collective licensing system for training data. Publishers get a cut, AI companies get legal cover. Messy but workable. Probably the most likely near-term outcome.
- Output rights get carved out. Courts hold that training is fair use, but AI output that directly substitutes for source material is not. Creates a two-tier system that will be a nightmare to enforce.
- Congress acts. Probably the least likely near-term outcome given how slowly IP law moves. But pressure is building, and there’s bipartisan motivation when the targets are both large tech companies and foreign competitors.
- Fair use expands and copyright atrophies. Courts decide transformation is so complete that most AI use qualifies, effectively gutting enforcement for a generation. Unlikely — but not impossible if lobbying balance tips hard enough.
The honest bottom line
Copyright was a bargain: society grants temporary monopoly rights to creators in exchange for eventual public domain contribution. AI breaks that bargain in both directions.
It learned from centuries of creative work without compensating anyone. It now produces work that may never need to enter the creative ecosystem at all. The legal system will patch something together — but it won’t be intellectually coherent. It’ll be whatever compromise the most powerful parties can negotiate.
The writers, photographers, and musicians are going to get something — probably not enough. The AI companies are going to pay something — probably not enough. That’s how these things usually resolve.
The people who built the internet learned this the hard way. Creators are learning it now.
The Hedge has covered financial and legal disruption since 2008. Brutal honesty over hype — always.
The Hedge 