Napster to Spotify: The Path AI Companies Should Copy

Everyone talks about disruption, but almost nobody copies the one playbook that actually worked. When Trip Adler looks at the chaos between AI companies and creators, he doesn’t see a new problem. He sees 2001 replaying itself.

The comparison is almost too clean: music in the Napster era was legally gray, freely accessed without permission, and defended by the argument that the technology was new and nobody knew the rules yet. Then iTunes and Spotify arrived and transformed the entire market by simply making the legal option more convenient than the pirate version.

Trip saw this before. At Scribd, he pioneered the subscription model for books. Now, building Created by Humans, he’s deliberately positioning it as the Spotify of AI licensing. Not because the analogy is clever, but because it’s predictive.

“Look back at the Napster era, right? Napster was also kind of legally gray area,” Trip explains. “And the answer to that was iTunes and Spotify, which is like building out the right, legitimate model. So that’s what we’re trying to be. We’re building out the iTunes and the Spotify of this space so that people don’t even want to use the Napster anymore. They want to just do the legit version.”

The pattern is powerful because it’s mechanistic, not moral. You don’t convince people to stop pirating by shaming them. You build something so convenient and accessible that the illegal option becomes friction.

The Napster Moment

In the early 2000s, music was a chaos of legal uncertainty. Napster let anyone share any song without paying rights holders. The labels sued. Napster lost. Then another file-sharing site appeared. Then another.

The music industry didn’t win that war with lawsuits. It won by accepting that distribution had fundamentally changed, and building a new business model around it.

Spotify’s stroke of genius wasn’t that it paid rights holders (they’d been paid before). It was that it made the paid version so frictionless that the paying option became the default for 99% of users. Picking a song was faster on Spotify than searching for a torrent. Offline listening, playlists, discovery algorithms — these weren’t magical. They were conveniences that made the legal option more attractive than piracy.

The AI and book world is in 2001. Companies are using data they didn’t pay for. The legal arguments are flying. Everyone is in “we don’t know if this is fair use yet” mode.

Trip isn’t betting the lawsuits will stop companies. He’s betting that once the legal landscape clears (through court rulings), the companies that already have legitimate relationships with creators will be faster to scale than the ones scrambling to retrofit legitimacy.

Why This Matters More for AI Than Music

The Napster comparison is tempting, but AI licensing is actually harder than music licensing was.

In music, rights holders are relatively concentrated. A label controls hundreds of songs. Spotify negotiated with a few major labels and most of the back catalog was covered. You didn’t need consent from every artist individually.

Books are fragmented. A single novel might involve an author, co-authors, a publisher, publishers in other territories, a literary agent, estate holders, and translators. An AI company that wants to license a million books faces millions of separate rights holders, each with their own stakes and concerns.

“They looked at how complicated the rights were and they’re like, this would be impossible to clear the rights for, because they all needed a lot of books, like a million books, it was just impossible to clear the rights for the books,” Trip says.

This fragmentation is why most AI companies haven’t even tried to build the “Spotify” layer. It’s why they’ve instead chosen to train on publicly available or scraped data and deal with the lawsuits later.

Created by Humans’ bet is that solving this fragmentation problem — building automation that lets individual authors and publishers claim their works and opt into licensing — is the Spotify play. It’s not about convincing companies to be moral. It’s about making the legitimate option less painful than the legal alternative.

The Three-Layer Model

Here’s how Trip frames the opportunity: the legal boundaries are still being drawn by courts. While they’re being drawn, companies need to decide which side of the line they’ll be on when it’s final.

“I think once these lawsuits mature, that’ll also be another catalyst for this to pick up even more,” he explains. “Once these lawsuits mature and courts draw the line.”

The companies that started licensing early — that said “we’re going to do this the legitimate way from the start” — won’t have the technical debt of retroactive negotiation. They won’t be explaining to a court why they used data they didn’t pay for. They’ll have clean chains of title.

The playbook is simple: move before you have to. Build the model now. When the courts make licensing mandatory, you’ll already be inside the system.

The Deal Reality Today

Trip is already doing this. Created by Humans has done deals with major AI companies, startups, even government agencies. But the volume is nowhere near what it needs to be. The real licensing boom hasn’t started yet.

He’s patient about this because he’s seen it before. At Scribd, it took years to convince publishers that digital subscriptions were viable. “I was there 17 years. It was actually a very steady grower over the years. If you look at the 17 year revenue growth curve, we grew every single year by, I think it’s like average of 30% growth per year and just kind of very steady upward growth for many years,” he reflects.

What changed wasn’t a sudden shift in publisher thinking. What changed was undeniable proof that digital worked. Created by Humans is betting on a similar pattern: the proof will come from court rulings saying licensing is required.

“One deal is one indication of the price and the second deal is another indication of the price and the third deal is another indication of the price and then after enough deals, you just get to the point where like, okay, well we’ve done 20 deals and this is kind of the average price, right?” he explains. “But I think we’re still kind of in that process of figuring that out.”

The Contingent Prize

The most interesting part of the Napster-to-Spotify comparison is that Spotify wasn’t built by the music industry. It was built by outsiders who understood both the technology and the business. Trip brings exactly that credibility to AI licensing — founder of a 16-year subscription business for books, now solving the fragmentation problem Spotify never had to solve.

But the real lesson from Spotify is this: once the legitimate option is convenient, demand moves. The question for AI companies isn’t “will we license?” It’s “will we be ready when licensing becomes inevitable?”

Trip is betting the answer is yes. And he’s building the infrastructure so that when the pivot happens, it doesn’t require negotiating with millions of individual authors from scratch. They’ll already be in the system.

FAQ

Why would AI companies wait for court rulings instead of licensing now?

Current legal uncertainty gives them a case for fair use. Once courts establish that unlicensed training is infringement, licensing becomes mandatory, not optional. Companies license today if they believe fair use will lose; they wait if they think they can win in court.

Is the Napster comparison actually accurate for AI?

Partially. Like Napster, AI training on unlicensed data is legally gray. Like Spotify, licensing could become the dominant model. But AI licensing is more complex than music licensing because books have multiple rights holders per title, not a single label per album.

Who has to move first for this to work — AI companies or creators?

Both, but the infrastructure has to exist first. Created by Humans is building the creator side (making it easy for authors to claim and license their work). Demand will come from AI companies once court rulings establish that licensing is required, not optional.

How long before AI licensing hits critical mass?

Trip’s experience suggests years, not months. Scribd took 17 years to reach scale. The AI licensing timeline will depend on how quickly courts resolve the pending lawsuits. Major rulings could accelerate adoption significantly.

Can an indie creator actually make money from AI licensing deals today?

Yes, but most of the volume today is from larger publishers with negotiating power. Smaller creators can opt into Created by Humans and license their work, but the deal volumes today are limited compared to what they’ll be once the legal landscape settles.

What happens if courts rule fair use does cover AI training?

That eliminates the licensing requirement, but it also doesn’t stop licensing from existing. Some companies might build licensing models for competitive reasons (better data, goodwill) even if it’s not legally required.

Why didn’t Amazon build this?

Amazon has publisher relationships and could theoretically build an AI licensing marketplace. Trip thinks they might — Amazon has the scale and relationships. But Created by Humans went direct to authors, which is a different bet: focusing on fragmented rights instead of publisher consolidation.

Could blockchain or smart contracts solve the rights fragmentation problem better?

Trip has thought about this extensively. Blockchain could theoretically distribute revenue automatically across multiple rights holders. But he chose a platform approach first (easier to iterate, more control) and plans to explore blockchain solutions after proving the core model works.

What’s a “reference deal” versus a “training deal” for AI?

Training deals let AI companies use books to train models from scratch. Reference deals let AI systems cite and retrieve from books (RAG). Transformative deals let AI create derivatives. Each has different legal and business risk, and authors are more comfortable with reference deals today.

How does Created by Humans actually ensure revenue gets split correctly?

The platform handles verification of rights ownership and automated revenue splitting. An author can claim a book, get verified as the rights holder, and set terms. When a deal is licensed, the platform enforces the split and distributes payment. This solves the fragmentation problem that made licensing look impossible.