AI Scores Court Wins—But the Copyright Battle’s Just Heating Up
Last week was a big one for AI companies—two major court wins in copyright cases gave them some breathing room in the legal chaos that’s been swirling since generative AI exploded onto the scene. In Bartz v. Anthropic and Kadrey v. Meta, federal judges ruled that these companies’ use of copyrighted materials to train their AI models was fair use. That’s a big deal.
But let’s not get carried away. The decisions are messy, complicated, and (shocker) don’t agree with each other on every point. Plus, they come on the heels of a February decision—Thomson Reuters v. ROSS Intelligence—that went the other way, saying nope, training AI on copyrighted materials wasn’t fair use. So it’s not exactly smooth sailing for OpenAI, Meta, Anthropic, and the rest of the AI crowd.
Here’s what’s going on: tons of lawsuits argue that AI companies broke the law by copying copyrighted stuff—books, news articles, art, you name it—to train their models. The companies say it’s fine because of the “fair use” doctrine, which is basically a legal safety net that says some copyright infringement is okay if the new use is transformative, doesn’t hurt the original market, and meets a few other factors.
But whether that argument holds up depends on the judge—and the facts.
In Anthropic, Judge Alsup said training Claude (Anthropic’s chatbot) on books was fair use because it’s like a human learning to write by reading. Even though Anthropic used full copies of copyrighted books, the court saw the AI’s output as something totally new, not a replacement. That said, Alsup drew a hard line: using pirated books was not fair use, period.
Over in Meta, Judge Chhabria also ruled in favor of the AI company, but took a different view. He didn’t care that Meta may have used pirated books—what mattered was whether Llama (Meta’s model) competed with or replaced the originals. Since it didn’t, the court gave Meta a pass. Chhabria did, however, float an interesting “what-if” argument: what if AI-generated books flood the market and drown out new authors? If the plaintiffs had made that case with solid evidence, he might’ve sent it to a jury. But they didn’t.
So where does this leave us? AI companies are winning—for now. But courts are split, and none of this is settled. These decisions could be appealed. Congress might jump in. And the next case could turn out totally differently based on the details.
Bottom line: fair use is still up for debate, and the copyright-AI showdown is just getting started.
Read more here: https://www.crowell.com/en/insights/client-alerts/ai-companies-prevail-in-path-breaking-decisions-on-fair-use