Poseidon, the data infrastructure built to source and license real-world data for AI, announced a partnership with Toss, the mobile financial platform operated by Viva Republica, to let everyday users contribute to AI training and be paid for what they provide. It is Toss's first move into AI data, opening that market to its roughly 30 million users.
The Specific Data Gap This Partnership Is Built to Fill
Frontier AI has run out of internet to scrape. The next generation of models depends on real-world data, the kind that captures how people speak, move, and react, which does not exist on the open web and has never had a clean way to be sourced, licensed, or paid for. The goal is building infrastructure to change that, with Toss bringing the reach to do it at scale.
What Users Actually Do Through This Partnership
Through the partnership, the contributor app, Numo, launches inside the Toss app directly. Toss users can help build Korean-language training data across voice, image, and video, and receive payment tied directly to what they contribute, rather than payment based on a flat rate disconnected from actual data quality or quantity provided.
What Makes the Data Numo Collects Specifically Valuable
What Numo collects is first-person data, recorded by real people in real environments, which is among the hardest and most valuable categories to obtain. It is the raw material for physical intelligence, the AI that has to operate in the physical world across robotics, autonomous vehicles, and other applications, distinct from the purely text-based training data that dominated earlier AI development.
Why Korea Specifically Was Chosen as the Starting Market
Demand from global AI labs for this kind of data is climbing, and Korea is positioned to supply it, with its dense real-life data and Toss's substantial user base. Poseidon and Toss intend to prove the model in Korea, then expand to global markets, treating this partnership as an initial proof of concept rather than the full scope of Poseidon's ambitions.
What Toss's Executive Director Said About This Move
Changhoon Seo, Executive Director of New Business at Toss, said as the AI industry grows, demand for high-quality data is rising just as fast. He framed Toss's plan as building an environment where users can take part in the data economy more easily and naturally, and expanding a structure in which the value users contribute is rewarded transparently.
How the Two Companies' Roles Actually Divide
Poseidon provides the infrastructure that tracks each contribution and its value, while Toss provides the user base and the financial experience that turns participation into actual payment. Together they offer a working answer to a question the AI industry has long struggled with: how to compensate the people whose data makes models better in the first place.
How Contributions Are Actually Verified and Tracked
Every contribution made through Numo is registered on DATA, the AI data network that Poseidon refines data for, giving each record a verifiable provenance trail through Trace, DATA's public audit layer, so a buyer can see where a piece of training data came from and a contributor can confirm their work was counted and paid.
What the Broader Backing and Track Record Look Like
A $15 million seed round was raised, led by Andreessen Horowitz (a16z). Numo has recorded more than 711,000 data registrations worldwide prior to this Toss integration, giving the partnership a base of existing platform traction rather than launching as a completely unproven concept.
Paying everyday users directly for contributing valuable AI training data like this reflects the same contributor-compensation model seen in The DATA Foundation's Trace provenance layer, both building infrastructure designed to properly source, verify, and compensate real-world data contributions at scale.
Glossary
- Physical intelligence: AI systems designed to operate and make decisions in the physical world, such as robotics and autonomous vehicles, requiring real-world sensory data.
- Contributor app: A platform through which individuals provide data in exchange for compensation, with contributions tracked and verified.
- First-person data: Data recorded directly by an individual in a real environment, capturing authentic human behavior rather than synthetic or scraped content.
Disclaimer
The information here is for general informational purposes only and does not amount to financial or investment advice. Data contribution platforms carry privacy, compensation, and market risk. Confirm current details directly through official Poseidon and Toss announcements.
