There's a gap in the plumbing of artificial intelligence, and it's about to get expensive.
Most of the world's commercially valuable data—the kind locked inside scientific journals, legal archives, financial terminals, and institutional repositories—remains out of reach for AI systems that can't or won't risk scraping it. With only about 1% of the world's data openly available for AI training, as Redpine estimates, that vast inaccessible information represents both a legal minefield and a business opportunity.
The Stockholm-based startup thinks it has found the seam. On April 28, 2026, Redpine closed an $8 million seed round led by NordicNinja VC to build what amounts to a licensing layer for the agentic economy. The funding, reported as €6.8 million in Europe, brings total capital raised to €9 million since the company's 2024 founding.
Luminar Ventures and node.vc joined the round. So did a notable roster of angels: Peter Sarlin, who co-founded Silo AI; Patrik Liu Tran from Validio; Anna Nordell Westling of Sana; and operators from OpenAI, Perplexity, and Spotify. The investor mix suggests belief in a thesis that the scraping economy—already under siege from copyright lawsuits—won't survive contact with regulation.
The Connective Tissue Problem
Redpine is building what it calls a "grounding API," though the label undersells the ambition. The platform acts as intermediary between rights holders and AI developers, offering API, MCP, and CLI access to premium content that's otherwise walled off. Developers pay per token consumed; content owners take a revenue share. It's infrastructure, not tooling—a headless backend that sidesteps the messier questions around fair use and transformative copying.
CEO Anders Hammarbäck, a former partner at Antler Sweden, has been blunt about the alternative. "A piracy-based scraping economy is only transient," he said in the funding announcement. His chief technology and product officer, David Österdahl—who put in early years at Spotify and iZettle before the latter's PayPal acquisition—echoed the framing: "Redpine is that solution."
Whether the market agrees is the gamble. Plenty of AI labs still scrape freely, and the legal landscape remains unsettled. But mounting copyright pressure from publishers and creators has made data provenance a reputational liability, if not yet a financial one.
What They're Building With the Money
The fresh capital will fund retrieval and reranking development, two unglamorous but critical capabilities for any system trying to surface the right needle in a haystack of licensed content. Redpine also plans to expand its partnership network and hire across engineering, data science, and go-to-market functions.
Early traction includes a January 2026 deal with AsedaSciences for scientific and clinical data integration, plus a customer relationship with Chegg. The company claims to be working with "leading international AI labs" but hasn't disclosed names—perhaps unsurprising given the political sensitivity around training data deals.
Marek Kiisa, the NordicNinja general partner joining Redpine's board, distilled the investment thesis to a single line: "Agentic AI is only as good as the data it's trained on." He added that most of the world's knowledge remains locked away from AI systems, a claim Redpine's own materials support with analysis pegging open data availability at roughly 1%.
A Different Kind of Data Play
The startup's positioning diverges sharply from annotation-focused companies like Scale AI, Appen, or Defined.ai. Those players help clean and label data; Redpine wants to be the access layer itself, with built-in attribution and provenance tracking that (theoretically) keeps everyone honest.
It's agent-native infrastructure, designed for a world where autonomous systems need to dynamically query premium sources rather than simply regurgitate pre-trained knowledge. Whether enterprises and developers will pay token-by-token for what some still acquire through automated scraping—well, that's the tension underpinning the entire round.
The company is betting that legal and ethical pressure will shift behavior faster than the technology community expects. If that calculus holds, Redpine could end up owning critical infrastructure in the agentic economy. If it doesn't, they've built expensive middleware for a problem the market hasn't yet agreed to solve.
