Redpine raises €6.8m to open up private data for the agentic AI era

Redpine raises €6.8m 

LLMs and AI systems have to train on vast quantities of data, but despite the sheer volume of content online and in print, just 1-2% of all data is publicly available for AI training.

Most AI startups and scaleups are working with the same, scraped internet data, often riddled with errors and hallucinations, but they are working with data that is in the public domain.

Stockholm-based Redpine is trying to help solve this problem and unlock the $200bm opportunity in agentic AI for tech service providers.

The startup, which has now raised a €6.8m seed round, is building a compliant marketplace that directly connects AI companies with content creators and data owners, opening up access to significantly more proprietary data.

The round, which follows a €1.1m pre-seed round back in September, was led by Finland-based NordicNinja, with participation from Stockholm-based Luminar Ventures and node.vc. Alongside the seed funding, Redpine has received investment from strategic angels including Peter Sarlin (co-founder SiloAI), Patrik Tran (co-founder Validio) and Anna Nordell Westling (co-founder Sana), and leaders from OpenAI, Perplexity, and Spotify. Nordic Ninja General Partner Marek Kiisa will also join the board.

Redpine was founded in 2024 by CEO Anders Hammarbäck, formerly head of Antler in the Nordics, and chief product and technology officer David Österdahl, a former venture partner at the same fund and an early Spotify team member.

The pair had known each other for eight years before deciding to tackle what they saw as one of the most overlooked layers in the AI stack: data.

“We felt like there was something missing here in the data layer of the AI stack,”  Hammarbäck told Pathfounders. “As an investor, I saw that the companies in the application layer and model layer we were looking at had very little differentiation as long as the data was all the same, so I was always looking for a data differentiation, and realised at some point we need to build that differentiation for other companies,”

That lack of differentiation is becoming a growing concern as AI companies increasingly build on similar foundation models and rely on the same publicly available datasets.

For Österdahl, the problem echoed an earlier technology shift.

Having helped build Spotify during its transition from piracy to licensed music streaming, he saw clear parallels in the current AI landscape, where creators are increasingly concerned about their work being used without permission. 

“There's a bunch of parallels with what's going on in AI right now and what was going on in music 20 years ago,” he told Pathfounders. “From that pattern recognition, we felt that we need to build a place where you can actually easily find the data you need to build the best possible service [...] but with data that makes sense to you and that didn't exist.”

Redpine’s platform is designed to make that process as seamless as streaming a song, allowing AI companies to license data directly from the source while ensuring that content owners are compensated every time their data is used.

There is, understandably, a lot of apprehension in the creative industries about content of all kinds being used to train AI without permission, with lawsuits popping up left, right and centre about YouTube videos, books and even song lyrics allegedly being used illegally. But the startup hopes to make things more equitable for creatives and operates on a pay-per-use model for both end users and creators, creating what it hopes will become an ongoing revenue stream for human content creators in the age of AI.

The company groups its datasets into three categories — “ground truth” (scientific data from research, non-fiction publications, journals and articles),  “real time signal” (data happening in real time such as the news and financial information) and “human taste” (restaurant reviews, magazine articles, content about media and arts etc).

Working with AI agent platforms building agents that need data for professional use cases, Redpine is particularly geared towards high-stakes sectors such as healthcare, finance, legal and manufacturing where there is no room for error or hallucinations.

The startup is already working with international AI labs and US-based biotechnology research firm AsedaSciences.

There are, naturally, other startups operating in this space as well, with the likes of Human Native AI and Created by Humans also vying to be the marketplace for creators to directly license their intellectual property, showing a broader shift towards licensed datasets as a source of differentiation.

Created by Humans, founded by the former CEO of Scribd, Trip Adler, attracted a similarly glittering lineup of angels, raising $5m seed from the likes of Craft Ventures founder David Sacks, Garry Tan and Floodgate Fund co-founder Mike Maples. The platform currently is tailored specifically towards authors and book publishers so isn’t competing directly with Redpine, but there are reportedly plans to expand to videos, images, music and medical data eventually. Human Native AI, meanwhile, was acquired by Cloudfare at the start of the year so has shifted direction.

“Agentic AI is only as good as the data it's trained on,” NordicNinja General Partner Marek Kiisa said. “Today, most of the world’s knowledge is still inaccessible to AI.”

The ambition, Hammarbäck says, is to build a company that makes a difference not only for powering AI — making the data trustworthy and reliable — but also to “nourish the roots of AI for the whole ecosystem to thrive”. 

“We decided this is the most important time to build a company that really matters for both the AI builders to make AI trustworthy and accurate, but also for the data and content owners to generate new, meaningful revenue streams in the age of AI agents,” Hammarbäck said. 

Redpine plans to use the funds to grow the team, speed up the development of the platform and  build more data partnerships to increase access to non-public datasets across industries.

Company information: