Startup tries to license AI authors’ work lawfully

Trip Adler’s startup, Created by Humans, is building a way for AI companies to use books with permission—after high-profile disputes showed how training datasets can become a battleground for creators’ rights.

On the night the debate felt like it was escalating, Trip Adler was already thinking about the same question that’s now haunting the AI industry: how do you stop creators’ work from being treated like free raw material?

Adler. the founder of Created by Humans. wants to put authors’ rights back into their own hands—and he’s doing it with a practical goal: letting AI companies use books. but only with permission from rights holders. The urgency is fueled by cases that have shown how quickly legal lines can blur when training data becomes an assumption.

In September 2025, Anthropic settled a class-action lawsuit over its use of pirated books to train its AI agent Claude. The payment was $1.5 billion, described as approximately $3,000 for each of the roughly 500,000 books included. In March 2026. journalist Julia Angwin filed a lawsuit against Grammarly—after its parent company rebranded itself as Superhuman—for mimicking Angwin and other creatives as part of its AI tool called “Expert Review.” That second lawsuit is still pending.

In both cases, the companies were accused of abusing authors’ rights. Anthropic’s matter was resolved in favor of the people whose works were taken. Superhuman’s case remains unresolved.

Adler says the core problem is the same: companies “used these books without permission.”

Created by Humans, launched with a different approach, aims to bridge the gap between creatives and AI—starting with the permission process. “For starters,” Adler says, he wants to “allow these companies to use the books, but with the permission of the rights holders.”

That sounds straightforward until you look at how messy copyright can be in practice.

As Adler puts it. “there’s not one simple way to clear the rights.” One reason. he said. is that “these AI rights have never been defined.” In his telling. the rights are “typically shared across multiple rights holders. ” including authors. publishers. and literary agents. Created by Humans, he says, works with authors and publishers who are open to AI. “We cleared the AI rights and then we get a licensing model going with AI companies.”.

Adler doesn’t come to this from theory. He described learning the realities of copyright the hard way, drawing on his earlier life in the book business through Scribd.

He said he had just graduated college when Scribd debuted in March 2007. “We had all these people uploading pirated books,” he said, “and that was how I got exposed to the book industry and how copyright works. I feel like you have to go down that path to really understand what copyright is about.”

Now, he’s trying to build a route that avoids repeating that kind of exposure for writers—by making it easier for AI companies to do the permission work instead of skipping it.

The timing matters because the rules are still being written.

The U.S. Copyright Office released part one of its Report on Copyright and Artificial Intelligence in July 2024, aimed at digital replicas. Part two, focused on copyrightability, came out in January 2025. The third part—addressing generative AI-training—was initially released in May 2025. but has not yet been published in its final form.

Within part two of that third document. section B is titled “Generative Learning Models. ” and it directly touches the kinds of disputes now playing out in the courts. The report says AI models “are well-known for requiring … millions or billions of works for training purposes.” It also notes that “text scraped from the internet often contains error messages or other content with limited or negative training value.”.

While regulators and courts work through these questions, publishing houses have been moving too.

In February, five of the largest publishers—Penguin Random House, HarperCollins, Simon & Schuster, Hachette, and Macmillan—announced a joint initiative aimed at protecting authors’ rights from AI.

Markus Dohle, CEO of Penguin Random House, said in a statement: “This is a watershed moment for the publishing industry. For the first time, we’re speaking with one voice on an issue that affects every writer we work with.”

The protections described by The Authors Manuscriptia include “mandatory disclosure when AI is used in any part of the book creation process. ” “explicit consent requirements before any author’s work can be used for AI training. ” and “a standardized compensation structure for authors whose works are included in AI training datasets.”.

Adler’s view is that the licensing challenge isn’t only legal—it’s cultural and practical.

He said it’s his belief “other people in Silicon Valley just don’t really understand” creative rights acquisition. “I think that’s kind of our role here,” he said, “to help people figure it out and just make it easier.”

His argument is blunt: AI companies don’t want to spend time negotiating an overwhelming number of individual deals. “People want to make it so that AI companies find this easy because they don’t want to go negotiate one million contracts with authors. They just can’t do that.”

There’s a clear tension running through the facts: while the U.S. Copyright Office is still moving toward final guidance, and major publishers are tightening consent and compensation, AI training continues to operate on assumptions that can land companies in court.

Without established rules dictating what AI can and cannot lift, more authors may find their work treated as takeable—while AI systems still move forward on the expectation that the raw material is already out there.

AI authors rights copyright generative AI licensing Anthropic Claude Superhuman Grammarly Expert Review Julia Angwin Created by Humans Scribd U.S. Copyright Office publishing industry Penguin Random House HarperCollins Simon & Schuster Hachette Macmillan