The AI copyright war: Why licensing may be our only chance at peace

Part I: Trouble in River City - The 1990s sampling wars

In the late 1980s and early 1990s, music faced an identity crisis. Hip-hop was exploding, and with it came the rise of "sampling," lifting snippets of existing recordings and weaving them into new songs. A drumbeat from James Brown, a bass line from Queen, a guitar lick from Led Zeppelin, these fragments became the raw materials of an entirely new musical language.

But there was a problem: nobody knew whether this was legal. Courts had barely begun to grapple with the question of whether a two-second riff or a four-bar loop counted as "copying" under copyright law. Artists were quick to complain when their works were used without permission, but many of those same artists were borrowing liberally themselves.

The result was chaos. Lawsuits piled up. Record labels scrambled to protect their catalogs. Hip-hop, once hailed as a new frontier of creativity, found itself in the middle of an industry-wide food fight.

One of the most famous early cases was Grand Upright Music, Ltd. v. Warner Bros. Records Inc. (780 F. Supp. 182 (S.D.N.Y. 1991)), involving rapper Biz Markie. He had used a portion of Gilbert O'Sullivan's "Alone Again (Naturally)" without permission. The court's ruling was blunt: "Thou shalt not steal." From that moment, the sampling landscape changed. What had been seen by many as an innovative art form was suddenly recast as outright theft in the eyes of the law.

Other cases followed. In Bridgeport Music, Inc. v. Dimension Films (410 F.3d 792 (6th Cir. 2005)), the court famously held "Get a license or do not sample," eliminating even the possibility of a de minimis use defense for digital samples. For many, this was the final nail in the coffin of the Wild West sampling era.

The message was clear: unlicensed sampling was infringement. The creative chaos of the 1990s had brought the industry to the brink, but litigation had provided no stable solution. What came next was not a courtroom victory, but industry adaptation.

Part II: Rationality returns - licensing systems take root

The breakthrough was not a single court case but a practical solution: licensing. Instead of endless lawsuits, artists and labels began to negotiate permissions for samples in advance. A market emerged where creativity and ownership could coexist.

Licensing wasn't perfect -- fees could be high, negotiations slow -- but it provided predictability. If you wanted to use a James Brown horn stab, you knew where to go, what the process was, and roughly what it would cost.

Some cases highlighted just how valuable these snippets could be. The drumbeat from The Winstons' "Amen Brother" became the most sampled rhythm in history, underpinning everything from hip-hop to jungle to pop. While the original creators never fully realized the value of their contribution, the controversy accelerated the industry's recognition that licensing was not just possible, but necessary.

By the 2000s, a functioning system had emerged. Artists still pushed boundaries, and lawsuits never disappeared, but there was now a mechanism to balance innovation with respect for ownership. Sampling became an accepted, indeed celebrated, part of music. The key was licensing.

Part III: Today's trouble in River City -- AI and copyright

Fast forward to 2025. Once again, creativity and technology are colliding and, once again, the law is scrambling to keep up. This time, the battleground is Artificial Intelligence.

Large language models and generative AI systems have been trained on massive datasets of human creativity, books, articles, songs, images and videos. They ingest, remix, and regurgitate material in ways that can be astonishingly powerful, but also unsettlingly derivative.

The parallels to sampling are obvious. Just as hip-hop producers once cut and looped recordings, AI tools now "sample" the world's creative works to generate text, music, and images. And just as in the 1990s, creators are asking: Is this theft or transformation?

The most high-profile battle so far is The New York Times Company v. Microsoft Corporation & OpenAI, Inc. (No. 1:2023cv11195, SDNY), filed in December 2023. The Times alleged that OpenAI used its content without permission to train ChatGPT and other AI models. According to the complaint, outputs from ChatGPT sometimes reproduced near-verbatim passages of Times articles, undermining the newspaper's subscription model.

OpenAI, for its part, argued for fair use. It claimed that its training methods were transformative, akin to a student reading widely before writing an essay. The company emphasized that its models didn't store or republish wholesale content, but instead generated new material based on patterns learned.

The stakes are enormous. On one side are the world's most powerful publishers, authors and media organizations. On the other are the technology companies driving the AI revolution. In between are creators of every kind -- journalists, musicians, academics -- wondering whether their work will be protected or consumed.

And much like the sampling wars of the 1990s, the likely outcome if left to litigation is years, perhaps decades, of uncertainty.

Part IV: A mediator's humble suggestion -- licensing AI

As a mediator, I can't help but see the parallels. Conflict thrives in ambiguity, and today's AI copyright disputes are rife with it. Courts can (and will) weigh in, but the result will almost certainly be piecemeal, inconsistent and slow. Meanwhile, the industry will stagnate under the weight of lawsuits.

The solution, as it was in music, is licensing.

Licensing transforms conflict into collaboration. It doesn't erase disagreement, but it channels it into agreements that both sides can live with. And critically, it provides a framework for innovation to continue without the constant threat of litigation.

How might licensing work in AI? Several approaches suggest themselves:

By length or extent

Just as music licensing sometimes considers how many seconds of a recording were used, AI licensing could account for the proportion of a work ingested or reproduced. Did the model scrape a single paragraph or an entire archive?

By evaluation of prominence

Not all works are equal. A throwaway news brief is different from a Pulitzer-winning investigative feature. Licensing could consider the prominence or distinctiveness of the material being used.

By artist status

In music, the estates of deceased artists often negotiate differently from living creators. Similarly, licensing for AI could distinguish between works by living authors (who rely on royalties) and historical archives.

By market impact

Courts already consider market harm in fair-use analysis. A licensing regime could formalize this, scaling fees based on whether AI use displaces the original market or creates new demand.

By collective management

Just as ASCAP and BMI manage performance rights for music, collective organizations could administer AI licensing. Publishers, authors, and artists could pool rights, enabling efficient negotiation rather than case-by-case litigation.

No system will be perfect. Just as music licensing did not eliminate disputes, AI licensing will not resolve every tension. But it offers a path forward that is fairer, faster, and more predictable than years of courtroom trench warfare.

Part V: Lessons from mediation -- from conflict to collaboration

At its heart, licensing is about balance. It recognizes that creators deserve recognition and compensation, while also acknowledging that new creativity often builds on the past.

As a mediator, I see this dynamic every day. Two parties come in ready for battle. Each believes they are right, and each fears losing ground. My role is not to declare a winner but to create space for resolution. Licensing is, in many ways, a mediator's solution writ large. It doesn't pretend to erase conflict, but it channels it into negotiated outcomes.

The alternative, endless litigation, serves few. Courts are slow. Outcomes are unpredictable. Relationships are destroyed -- and innovation stalls under the weight of fear.

We've been here before. The music industry of the 1990s shows us what happens when technology disrupts art without a clear framework. At first, chaos. Then, lawsuits. And finally, a recognition that licensing provides the only sustainable path forward.

AI is no different. The genie is out of the bottle. The question is not whether we can stop it, but whether we can create systems that allow innovation to thrive while honoring the rights of creators.

Conclusion: finding harmony

The great irony of the sampling wars is that they ultimately enriched music. Licensing didn't stifle creativity; it enabled it. Hip-hop, electronic and pop music flourished once artists knew the rules of the game. Creativity and commerce found harmony.

We stand at a similar crossroads with AI. We can spend the next two decades fighting in court, or we can learn from history. Licensing offers a model for moving forward. It won't solve everything, but it will provide predictability, fairness and space for innovation.

As a mediator, I find that deeply appealing. If music, an art form built on repetition and reinvention, could find harmony through licensing, then so too can AI. The alternative is constant litigation. Nobody benefits when the fighting never stops. In the end, progress doesn't come from endless fighting, it comes from finding common ground.