It’s our pleasure to introduce you to Jason Kelly, our newest storyteller at Unicorner. And he brings to you, Created by Humans. In the age of AI, one of the greatest controversies continues to surround how to license human-made data used to train billion-dollar LLMs. Created by Humans is trying to build the answer. Enjoy!

In case you missed it: Tech Week is around the corner. We have many events coming your way in both SF and LA. Just to get a sense of who we can expect, give us a light RSVP so we can clear the room!

Ethan and Arek 🦄

Authors can spend anywhere from months to years writing their books. What if you put all that effort into writing a book, only to discover an AI company scanned every word to train a multi-billion-dollar model without asking permission or paying you? That has been the reality for most authors since generative AI became the cultural zeitgeist, and until recently, a slow-moving lawsuit was the only leverage they had to do anything about it.

Created by Humans believes that when an AI learns from human creativity, that relationship must be symbiotic and compensated. To enforce this, it has built a licensing marketplace that allows authors, agents, and publishers to dictate exactly how AI labs can use their books and collect payments for creators when they do. The company handles the legal and technical work of turning those rights into different types of deals, depending on what an AI company wants to do with the content.

The goal is to give AI companies a clean, legal way to access high-quality writing while ensuring the people who actually wrote it keep getting paid for it.

Check it out: createdbyhumans.ai

Created by Humans keeps a percentage of the resulting licensing revenue as a service fee, passing the majority back to the rights holder. The service fee varies according to the kind of deal.

Training deals are upfront, lump-sum payments for annual or perpetual access to a body of work, meant to allow an AI model to learn from content, but not directly reproduce it.

Reference deals apply when a model cites or pulls from a specific book in an actual response, and rights holders get paid on a per-use basis rather than a one-time sale.

Transformative deals cover cases where an AI company builds something entirely new from a licensed work, and those terms get negotiated case by case.

After spending almost two decades working in the publishing industry, Trip Adler founded Created by Humans, having realized that the internet may change how people access written work, but it shouldn’t destroy the value of people’s creativity. 

Adler co-founded Scribd out of Harvard in 2007. He ran the company as CEO for roughly 15 years, steering it into one of the largest reading and digital document subscription platforms in the world before stepping down in 2023. Through that tenure, he accumulated institutional relationships with major authors, literary agents, and legacy publishing houses. When generative AI models began aggressively scraping the open web to train large language models, Adler had a front-row seat to the publishing industry's reaction. Creators may not be as afraid of the software itself as they are angry about being excluded from the financial upside of the AI boom. 

To bridge this divide, Adler recruited two former partners from his Scribd days: Edward Igushev to architect the technical platform and Jen Singerman to lead marketing and operations. The team saw an opportunity to build a legal and technical framework to hold AI labs accountable for the information and creativity they were siphoning out of the market. That framework is what the team calls "the Fourth Law.”

Every LLM learned how to write by reading things they didn’t pay for.

For years, that was simply how AI companies operated, because they scrape the open internet and sort out the legal questions later. Writers noticed. The Authors Guild sued OpenAI over it, arguing that training a model on someone's life's work without a license is copyright infringement. As AI models evolve every few months, waiting for a resolution to go through the legal system is a futile defense while the technology continues to benefit from unpaid artists’ work. Cases like this can drag on for years, and the model in question may look nothing like it did when the suit was filed.

Created by Humans is betting there's a faster way to get authors paid than waiting on a judge. The startup is building the pipeline for the deals that come next by clearing rights, converting a manuscript into something an AI company can use, and negotiating on the author's behalf. A single training license can bring in several thousand dollars per title, split between author and publisher, according to the Authors Guild's own breakdown of the platform.

That's money for a group of people who made nothing from it a few years ago.

It's also more than a story about books. Created by Humans has said it eventually wants to license video, images, music, and even medical data the same way, which would make it something closer to a licensing utility for the whole creative economy.

Whether enough authors trust the system to actually use it is still an open question. When NPR covered the platform’s launch, only about a third of the authors surveyed said they'd be willing to license their work to an AI company at all, and Susan Orlean (author and staff writer for The New Yorker) called the protections on offer "conceptual rather than actual" until money actually changes hands. Created by Humans can prove the naysayers wrong by continuing to accrue paid deals.

Not everyone needs convincing. Walter Isaacson, a well-known author on the platform, invested in the company and signed on as a rights holder, and has said Created by Humans is "solving authors' key challenges with AI: control over how our work is used by AI companies, the accuracy of outputs referencing our work, and getting compensation for our work's use."

The publishing industry is already leaning in this direction. Microsoft struck its own licensing deal with HarperCollins in late 2024, paying $5,000 per nonfiction title used for AI training. This is proof that a major AI lab will pay for exactly the kind of content Created by Humans is aggregating. Research from Epoch AI, a company that is investigating the trajectory of AI for the benefit of society, suggests the open internet could run out of fresh, high-quality, unscraped text sometime between 2026 and 2032, which only makes licensed, legally clean writing more valuable over time.

If Created by Humans can turn that scarcity into leverage for the authors who wrote the words, it has a good shot at becoming the tollbooth the AI industry has to pass through. It seems that the Fourth Law of Robotics has a real chance at becoming a reality thanks to the work Adler and his team are doing.