When Cashmere.io announced it had closed a $5 million seed round led by Reach Capital, with participation from Ingram Content Group, Pearson, and Naver, few in the publishing industry had heard of the company. That was by design.

“We’ve been very deliberate,” said Jonathan Munk, Cashmere’s cofounder and CEO. “We want to build with partners who believe what we believe and see what we see and want to build alongside us.”

The Salt Lake City–based startup has built an infrastructure layer that enables publishers to license, monitor, and monetize their content within AI systems. The company also received a $1 million pre-seed investment from Perplexity, the AI search platform that became its first client—and, as it turns out, has been using Cashmere’s technology for some time.

The funding underscores a shift in how publishers are approaching the AI economy—moving beyond using unauthorized scraping toward building commercial infrastructure that lets them set terms, track usage, and get paid. With the global publishing market valued at over $150 billion, the race to build the licensing layer connecting publisher content to AI platforms is intensifying.

Munk came to publishing from the outside. He spent 11 years in EdTech, building software for leadership development and corporate training, before being brought into a company called Book Club about three years ago that licensed content from authors and publishers to create training programs. It was there that he and his cofounders identified what they saw as a gap: Publishers had no adequate tools for managing how their content was being consumed by emerging AI technologies.

“Publishers want and need better tools to be able to bring their content into AI,” Munk said. “The question is, what’s the legal, governed, managed, high-quality, commercially viable way for the content to live? We’re essentially creating the rails for that transformation.”

As opposed to what might happen in a bulk catalog licensing deal, Cashmere’s technology ingests publisher content and parses it into a proprietary knowledge graph format the company calls an “omnipub,” which structures unstructured data for AI consumption.

“If you were to feed in Seven Habits of Highly Effective People as a book into AI, it wouldn’t know where one habit starts and one habit stops,” Munk said.

Once structured, publishers and the AI alike can manipulate the relationship more effectively. Content is licensed using per-token, per-use pricing, or on a per-relationship basis, and publishers have control over granting and rescinding licenses. A dashboard allows them to track consumption and define or redefine the terms of their licenses based on data.

A core element of Cashmere’s pitch is IP protection, and the company promises its technology never exposes a publisher’s full content to an AI platform. “It’s near impossible to reconstruct a book or a catalog or an article because of the way that we’re architected,” Munk said. “We’re delivering just little snippets of content on a per-use basis rather than full text.”

Cashmere built the integration that powered the Wiley-Perplexity content licensing partnership and now handles all of Perplexity’s premium data integrations.

“Basically anytime that Perplexity brings in a new data partner, they run it on Cashmere,” Munk said. The company is also working with Harvard Business Publishing, according to Munk, with additional partnerships in the pipeline.

At present, Cashmere has fewer than 10 employees and plans to stay lean. Munk said both rounds of funding were oversubscribed and he has no plans to accelerate growth.

Munk drew a distinction between Cashmere’s approach and the model context protocol, or MCP—an open standard developed by Anthropic that allows AI models to connect directly to external data sources and tools. MCPs have gained significant traction in the tech world as a way for AI systems to pull in outside information on the fly. But Munk argued that for publishers, the protocol lacks the controls that matter most.

“It’s quite easy these days to just build an MCP,” he said. “But the question is, what controls sit behind that MCP? What management, where does the data go? What parameters can you put on it? How do you track it? How do you limit access of the LLM to the data?”

Put another way: “MCPs are a blunt instrument,” Munk said. “The question, if you’re a publisher, is to ask, what needs to be true for us to feel good about bringing our content and selling our content into AI systems?”

The answer, he argued, requires layered controls around compensation, credit, and content management—precisely the infrastructure he claims Cashmere has built.

Drawing a parallel to the music industry’s transformation after Napster, Munk acknowledged that widespread unauthorized scraping has already occurred.

“Yes, everyone has the content already,” he said. “The question is quite simple: what’s the legal, governed, managed, high-quality, commercially viable way for the content to live. We aim to answer that.”