By Matthew B. Harrison
TALKERS, VP/Associate Publisher
Harrison Media Law, Senior Partner
Goodphone Communications, Executive Producer
Every media creator knows this moment. You are building a segment, you find the clip that makes the point land, and then the hesitation kicks in. Can I use this? Or am I about to invite a problem that distracts from the work itself?
That question has always lived at the center of fair use. What has changed is not the question, but the context around it. Over the past year, two federal court decisions involving AI training have quietly clarified how judges are thinking about copying, transformation, and risk in a media environment that looks nothing like the one for which these rules were originally written.
Fair use was never meant to be static. Anyone treating it as a checklist with guaranteed outcomes is working from an outdated playbook. What we actually have is a 20th century rulebook being used to officiate a game that keeps inventing new positions mid-play. The rules still apply. But how they are interpreted depends heavily on what the technology is doing and why.
That tension showed up clearly in two cases out of the Northern District of California last summer. In both, the courts addressed whether training AI systems on copyrighted books could qualify as fair use. These were not headline-grabbing decisions, but they mattered. The judges declined to declare AI training inherently illegal. At the same time, they refused to give it a free pass.
What drove the analysis was context. What material was used. How it was ingested. What the system produced afterward. And, critically, whether the output functioned as a replacement for the original works or something meaningfully different. Reading the opinions, you get the sense that the courts are no longer talking about “AI” as a single concept. Each model is treated almost as its own actor, with its own risk profile.
A simple medical analogy helps. Two patients can take the same medication and have very different outcomes. Dosage matters. Chemistry matters. Timing matters. Courts are beginning to approach AI the same way. The same training data does not guarantee the same behavior, and fair use analysis has to account for that reality.
So why should this matter to someone deciding whether to play a 22-second news clip?
Because the courts relied on the same four factors that govern traditional media use. Purpose. Nature. Amount. Market effect. They did not invent a new test for AI. They applied the existing one with a sharper focus on transformation and substitution. That tells us something important. The framework has not changed. The scrutiny has.
Once you see that, everyday editorial decisions become easier to evaluate. Commentary versus duplication. Reporting versus repackaging. Illustration versus substitution. These are not abstract legal concepts. They are practical distinctions creators make every day, often instinctively. The courts are signaling that those instincts still matter, but they need to be exercised with awareness, not habit.
The mistake I see most often is treating fair use as permission rather than analysis. Fair use is not a shield you invoke after the fact. It is a lens you apply before you hit publish. The recent AI cases reinforce that point. Judges are not interested in labels. They are interested in function and effect.
Fair use has always evolved alongside technology. Printing presses, photocopiers, home recording, digital editing, streaming. AI is just the newest stress test. The takeaway is not panic, and it is not complacency. It is attention.
If you work in the media today, the smart move is to understand how the rulebook is being interpreted while you are busy playing the game. The rules still count. The field just looks different now.
Matthew B. Harrison is a media and intellectual property attorney who advises radio hosts, content creators, and creative entrepreneurs. He has written extensively on fair use, AI law, and the future of digital rights. Reach him at Matthew@HarrisonMediaLaw.com or read more at TALKERS.com.