Palo Alto, United States, June 26th, 2026 - Poseidon, a data infrastructure company focused on sourcing and licensing real-world data for artificial intelligence, has formed a partnership with Toss, the mobile financial platform run by Viva Republica, to enable Toss users to supply and be paid for training data. The initiative launches first in Korea and opens the data economy to Toss’s roughly 30 million users.
Working through the Toss app, Poseidon’s contributor application, Numo, will let everyday users produce and register first-person recordings in Korean across voice, image and video. Poseidon says the system ties payments directly to the contributions that users make, and it supplies the underlying tracking and valuation infrastructure while Toss supplies the user base and the payment experience.
Poseidon positions the offering as an answer to a growing shortfall in training data for next-generation AI systems. The company argues that frontier models cannot rely solely on publicly available web content and instead need real-world data that reflects how people actually speak, move and react in everyday environments. According to Poseidon, that kind of material is among the most valuable and hardest-to-source categories for model builders.
Every submission captured through Numo is recorded on DATA, the AI data network that Poseidon refines data for. DATA provides a verifiable provenance trail through Trace, its public audit layer, enabling buyers to see the origin of specific training records and contributors to confirm that their contributions were counted and compensated. DATA Foundation, which launched this week after rebranding from Story, is developing this layer with integration partners including the human data marketplace Kled, and Poseidon is listed as one of the largest suppliers of refined data flowing into the network.
Poseidon describes the collected material as "first-person data," recorded by individuals in real settings. The company views this as the raw material for so-called physical intelligence - the class of AI that must operate in the physical world across robotics, autonomous vehicles and related applications. Demand from global AI labs for these categories of data is rising, the partners say, and Korea offers a concentrated source of such real-life data together with Toss’s large user base.
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. Toss plans to build an environment where users can take part in the data economy more easily and naturally, and to expand a structure in which the value they contribute is rewarded transparently."
SY Lee, Chief Strategy Officer and Chairman of Poseidon, added: "Korea is one of the few markets where the strategic importance of AI data, a mature financial system, and world-class mobile experience all exist at once. Toss is the right partner to turn user-contributed AI data from an early idea into a standard the rest of the world can adopt." The article notes Lee’s past roles founding the web-novel platform Radish and co-founding Story, and that he was named a Young Global Leader by the World Economic Forum this year.
Poseidon’s contributor app, Numo, has already recorded more than 711,000 data registrations worldwide and is now available inside Toss. The company says it was incubated by the team behind The DATA Network and builds a bridge between data supply and AI demand by creating access to IP-safe, composable training datasets. Poseidon also previously raised a $15 million seed round led by Andreessen Horowitz.
The partnership is framed as a pilot to prove a compensation and provenance model for user-provided AI training material in Korea before scaling globally. Toss contributes the payment rails and large user base, while Poseidon supplies the instrumentation and registration infrastructure that tracks contribution provenance and value for buyers and contributors alike.
Contextual analysis
For companies building AI that must operate in the physical world, obtaining labeled, real-world recordings is essential and often more costly than scraping public text. The Toss-Poseidon deal combines a high-reach mobile platform and embedded payments with an auditable data registry, creating a linked workflow from data capture through valuation to payout. If the model scales, it could alter how human-generated training datasets are sourced, licensed and monetized - particularly in markets where mobile penetration and payment systems are advanced.
Operational highlights
- Numo is embedded inside the Toss app to collect first-person voice, image and video data in Korean.
- All contributions are recorded on DATA and traceable via the Trace audit layer for provenance.
- Poseidon provides valuation and registration infrastructure while Toss handles user reach and payments.
- Poseidon reports more than 711,000 Numo registrations globally and a $15 million seed round led by a16z.
Next steps outlined by the partners
The firms say they will validate the model in Korea and then pursue international expansion. The partners framed Korea’s ecosystem - a combination of concentrated real-life data, a mature financial system and strong mobile experiences - as a strategic testing ground for a model they hope can become a broader industry standard.
Reporter
Leila Farooq covers wireless carriers, broadband and streaming ecosystems with a focus on subscriber quality, pricing discipline, and capital allocation tradeoffs. Her reporting here examines how a mobile financial platform can be leveraged to create a paid contributor workflow for AI data.