When Should Your AI Startup Actually Start Licensing Data?

The founder dilemma is as old as startups: move fast or move right?

Most founders in the AI space have chosen to move fast. They’ve trained on whatever data is available, relied on fair use arguments, and assumed that if they get big enough, they’ll figure out the legal questions later. Some of that will work out. Some won’t.

But Trip Adler is telling founders something different: the cost of moving right has gotten a lot cheaper. And the cost of moving fast is about to get a lot more expensive.

His pitch is simple: licensing is no longer the expensive, complicated option. Created by Humans has made it operationally straightforward. And once courts rule on fair use, the companies that didn’t license will be in retroactive negotiation mode from day one.

“If you’re a startup founder listening to this and you’re looking for licensed data, I mean, we’d love to talk,” Trip says. “Because I think it’s the right way to do things. And it’s a great way to just get visibility and do things the right way.”

But he’s realistic about why startups don’t. Founders want to prove product-market fit before they complicate things with licensing. They want to move fast. They don’t want lawyers involved.

So the real question is: at what point does the cost of moving fast exceed the cost of moving right?

The Founder Playbook: How Most Do It

Trip has lived this twice — once as a confused founder at Scribd, once as a founder who understands it at Created by Humans. His honesty about the Scribd story shows why most founders move fast first.

“I mean, I could tell you like in my last company, Scribd, I mean, I’ll give a little background how I got into the world of copyright. So when I was just out of college, I launched the site Scribd to let people upload documents to the web. And I didn’t at the time know anything about copyright.”

What happened next was inevitable: users uploaded pirated content. Publishing companies sued. Scribd had to build copyright compliance from scratch, while the company was growing and serving paying customers.

This is the standard founder journey. You build, users do things you didn’t anticipate, you react, you learn, you build compliance around it. It’s not evil. It’s just the cost of moving fast in legally gray territory.

But Trip learned something from that experience that’s relevant now: “I think that’s a common founder journey. But I think ultimately though, I think like, you know, it’s the earlier you embrace copyright, the better.”

The Changed Calculation

What’s different in 2026 versus 2007 is this: licensing is no longer operationally impossible.

“I mean, if you’re a startup founder listening to this and you’re looking for licensed data, we’d love to talk,” Trip offers. “We structure all kinds of deals. It could be a deal with a large upfront payment or it could be a deal that’s entirely usage-based. So yeah, we want to work with companies of all sizes and if you’re a small team and you want to try to do this, we’re very happy to work on it with you and find a model that works.”

The licensing models work for startups now:

• Usage-based: Pay only for the data you actually use. No upfront cost. As you scale, licensing costs scale.
• Upfront payment: Pay once, get access. Good for companies with funding.
• Revenue share: Some percentage of your AI revenue goes to creators. Aligns incentives.

The second change is legal precedent. Anthropic’s $1.5B settlement, Britannica suing OpenAI, the Merriam-Webster cases — they’re drawing the line on what fair use covers.

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

So the calculation changes. Moving fast now means:

• Upside: Get to market first, prove product-market fit without legal overhead
• Downside: Once courts rule, you have to retroactively license, negotiate with angry creators, possibly defend past usage

Moving right now means:

• Upside: When courts rule licensing is required, you’re already in compliance, your path to scale is clear
• Downside: Legal overhead now, licensing costs now, maybe slower to first revenue

For most founders, moving fast still wins. But the math is closer than it was.

The Real Cost of Moving Fast

Trip points out something most founders miss: the cost of licensing today is not the cost you’ll pay forever.

“I mean, think ultimately if you get deals done, 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.

The market is still discovering pricing. This means licensing might be expensive today and cheap tomorrow. Or cheap today and expensive tomorrow. You won’t know until the market settles.

But you will know: companies that licensed early will have better relationships and clearer title when pricing settles. Companies that delayed will be negotiating at a disadvantage.

The Gray Area Founders Live In

Trip doesn’t pretend there’s a clear answer. He just acknowledges the trade-off and lets founders choose.

“A lot of founders just want to move fast. And they don’t think about this question very much. And that’s why there is just a lot of piracy going on in the world of AI and it’s all kind of legally gray area,” he notes.

This is where his Scribd lesson applies. He moved fast, hit legal problems, learned compliance, and then figured out how to do it right. That journey worked for Scribd. It might work for you.

But the difference now is the decision is visible. You can choose it consciously instead of stumbling into it.

“But I think over time we’ll kind of build new standards here, build new solutions and clear the rights. And eventually over time, this will become just a more like legally clean situation,” he predicts.

What Actually Moves a Startup to License

Trip’s deal experience shows what makes a founder actually move toward licensing:

1. You want to build goodwill — especially if you’re in a space where your user base includes creators (content platforms, creative tools)

2. You want cleaner signal — licensing data you paid for is better signal for model quality than pirated data. Investors notice this.

3. Your model requires trust — if you’re selling to enterprises or regulated industries, having clean licensing matters immediately, not later.

4. You want visibility — licensing deals with major publishers and platforms create credibility signals that investors and partners recognize.

Trip emphasizes this last one: “Because I think it’s the right way to do things. And it’s a great way to just get visibility and do things the right way.”

For most startup founders, this won’t move the needle. They’re focused on the next fundraising round, not on creator relationships or investor signal. They’ll move fast.

But for founders building in spaces where licensing is already expected (creator tools, media, enterprise software), the calculus is different.

The Inevitable Transition

What Trip is betting on is this: the founders who move fast now will eventually have to license. The only question is when and at what cost.

“But we think over time we’ll kind of smooth it out and get everyone working together,” he says.

The companies that will scale fastest in the long term are the ones that realize this early and build licensing into their model before it’s mandatory. Not because it’s morally right (though Trip believes it is), but because it’s strategically smarter.

“I think ultimately though, I think like, you know, it’s the earlier you embrace copyright, the better,” he concludes.

That’s not a moral argument. That’s a business one.

FAQ

Should I wait for court rulings before starting to license?

No. Start the conversation now, structure deals that work for your stage, and learn the landscape. Waiting until courts rule puts you behind companies that already have relationships and know what works.

How much do licensing deals actually cost for startups?

Depends entirely on the deal. Usage-based deals scale with your usage (could be very cheap early). Upfront deals depend on negotiation. Revenue share aligns incentives but requires profitable models. Trip structures deals for companies of all sizes.

What if I use open or public data instead?

That’s smart for some use cases (research, synthetic data). But if you’re training on books, papers, or content that has clear creators, you’ll eventually hit questions about fair use. Public doesn’t equal “free to train on without permission.”

Can I license data retroactively after I’m already trained on it?

Technically yes, but it’s weaker positioning. You’ve already demonstrated you don’t think licensing is required. Created by Humans will work with you anyway, but it’s easier if you license first.

What’s the difference between fair use and licensing?

Fair use is a legal argument that lets you use copyrighted material in certain contexts (criticism, parody, education, research). Licensing is paying for explicit permission. Courts are currently deciding whether AI training qualifies as fair use.

How do I know if my AI product needs data licensing?

If you’re training on copyrighted content (books, articles, music, video) that has clear creators, you eventually need licensing or legal certainty. If you’re training on public data, research papers, or synthetic data, you might not. But “might” is the legal gray area.

Should I have lawyers review licensing before I sign?

Yes. Trip emphasizes he can’t encourage copyright infringement. Licensing is a legal agreement. You should understand it. For startups, this might mean splitting a lawyer with other founders or using legal templates, but don’t skip this step.

What if courts rule that AI training is fair use?

Then licensing becomes optional instead of required. But it doesn’t stop being valuable — companies can license for competitive reasons (better data, creator goodwill, investor signal). Either way, having licensing relationships gives you optionality.

Can I structure a licensing deal to be cheap initially and scale later?

Yes. Usage-based deals let you pay as you grow. You’re not paying for licenses you don’t use. Created by Humans works with all deal structures, so ask what models work for your business stage.

What happens if I’m wrong about fair use and courts rule against me?

You owe damages and have to stop using the data. This is why licensing early is cheaper — you’re paying fair market price now instead of damages plus legal costs later.

How do I actually talk to Created by Humans about a deal?

Trip says founders should reach out at any stage. You don’t need a polished pitch. You need a clear statement of what data you want to use and why. The team will talk through deal structures that work for your situation.