"Fair Use" Fails! First Court Ruling Is A Game-Changer for Artists, Media & GenAI Tech

I've Long Predicted Courts Will Ultimately Reject Tech's Fundamental Defense for Training LLMs Without Consent & Compensation. And Now, The First One Has.

Welcome to an emergency “bombshell edition” of the brAIn! Yesterday, in the closely watched AI infringement case Thomson Reuters v. ROSS Intelligence, a federal judge ruled for the first time on the fundamental issue of “fair use” in the context of AI training on copyrighted works — i.e., whether training without consent and compensation is infringement or, instead, a defensible fair use. And Judge Stephanos Bibas’s decision was a resounding “yes” to infringement as a matter of law, which means a resounding victory for media, entertainment, artists and the entire creative community — and aggressive, stinging loss for Silicon Valley’s singular refrain of “fair use.”

The Court’s decision is one I’ve long predicted (read my earlier analysis here about how the Supreme Court would be likely to decide) — and one that I’ve always argued as being the right one both as a matter of law and ethics.

To be clear, this is just the first of many court rulings on the issue of “fair use,” since 40+ relevant litigations continue to wind their ways through the U.S. courts. But the first decision is always the most significant one — with out-sized impacts — as it sets the tone for the other decisions to come. And certainly many more are coming soon later this year.

The losing defendants here — aka “the Infringers” — will most certainly appeal the Court’s ruling, which the Judge made alone on Summary Judgment pre-trial. So this game certainly ain’t over yet.

But, no matter how you slice it (and many will try to slice it the other way), it’s a huge win for media and entertainment. Their stock is now (likely literally) on the rise. And the Court’s decision represents a significant blow to all generative AI developers that have scraped and scraped copyrighted content for LLM training for years with casual indifference and seeming impunity. Now there is very significant potential “punity.”

Ultimately, the Court’s rationale for rejecting “fair use” was primarily based on a finding that the defendant, Ross Intelligence, had used plaintiff Thompson Reuters’ copyrighted works for training its AI in order to compete directly with Thompson Reuters. In other words, “market substitution” — direct predatory commercial harm.

Those critical words — “market substitution” — are the words and rationale that formed the basis for the U.S. Supreme Court’s most recent rejection of “fair use” as a defense in the famous Andy Warhol / Prince” case, about which I’ve written several times. That is significant “symmetry” in reasoning, in fact. Judge Bibas’s words here in Thompson Reuters are no accident. They parrot the words of the Supreme Court. And that bodes well for the creative community for future decisions by other courts. Legal parroting leads to more parroting. And ultimately, if any one of these cases make it all the way to the Supreme Court, those nine wacky souls on the bench will have their own earlier words and “market substitution” rationale to fall back on. And that means that GenAI developers have much to worry about. And the creative community much to cheer.

This new decision — and blunt force trauma rejection of “fair use” as a defense to wholesale “taking” — are likely to have immediate significant ramifications. Dynamics between the media companies that create content — and the generative AI developers that need that quality media content to feed their LLMs — have now immediately shifted. While media can’t control the pace of tech’s AI development, media now can wield the power to place very real fear (potential massive legal and financial exposure) into the hearts of those generative AI developers that continue with a “damn the torpedos” approach.

One immediate impact? I predict a frenzy of major new content licensing activity and deals between media and tech — accompanied by the significantly higher licensing fees that should flow from spiked demand.

And isn’t that the right answer? I’ve always preached “win win” cooperation rather than litigation. It’s only fair — and smart — to pay content owners for the value they bring to the generative AI equation. Fair dealmaking erases the friction of risk, acrimony and litigation — and eradicating friction is the best way to spur massive AI adoption.

In other words, Judge Bibas got it right.

Reach out to me with your feedback to peter@creativemedia.biz.