Origin Lab raises $8 million to connect video game studios with AI labs seeking world-model training data

Origin Lab, a newly funded startup, is positioning itself as a data licensing intermediary in the emerging world-model AI sector. The company closed an $8 million seed round with backing from Lightspeed Ventures alongside several venture firms and high-profile angel investors from the gaming and autonomous vehicle industries.

The platform targets a specific technical gap: AI labs building models of physics and spatial dynamics require large volumes of realistic video footage and environmental interaction data. Unlike text corpora for large language models, such training material has historically been scraped from video games and streaming platforms without explicit licensing agreements, creating legal and quality friction.

According to co-CEO Anne-Margot Rodde, the company will handle the technical work of converting game assets into usable training formats—ranging from simple rendering passes to automated extraction from gameplay footage. On the supply side, gaming studios gain new revenue streams from existing digital assets; on the demand side, research organizations gain legal clarity and curated data quality.

The timing reflects broader industry challenges. OpenAI's Sora model drew scrutiny in late 2024 for apparently reproducing game footage and streamer content, highlighting the unlicensed training data problem. Amazon has separately acknowledged its interest in Twitch as a training resource, underscoring demand for this category of data.

Lightspeed's investment thesis centers on the analogy to data-vendor success stories. Scale AI's rapid scaling has validated the market for companies that supply curated training material to capital-intensive AI labs, where data acquisition remains a primary bottleneck rather than a technical challenge.