Data Marketplaces & Intermediaries

Protege partners with Dailymotion to license ~40M videos for AI training

Source: Protege · Jul 28, 2026

Video is the training-data category everyone in AI acknowledges is scarce and almost nobody has solved cleanly at scale — rights-protected, high-quality, globally diverse video is genuinely hard to source compared to text or even audio. Protege's new partnership with Dailymotion is a direct attempt to close part of that gap: the deal makes roughly 40 million videos from Dailymotion's multilingual catalog available to AI model builders through Protege's data platform.

Protege itself operates as an intermediary — it doesn't own the underlying content, but instead runs the data platform layer that helps content owners like Dailymotion monetize their libraries for AI training while retaining control over how that content is used. The company's stated role covers dataset curation, preparation, and research support, positioning it as connective infrastructure between content owners with valuable libraries and AI developers who need licensed access to that content at scale.

The announcement leans heavily on the language of responsible licensing — a "licensed, transparent framework" and "rights-cleared access" are both explicitly emphasized — which reads as a direct response to the broader industry criticism that most large-scale AI video training has historically relied on scraped or ambiguously-licensed content. Notably, the announcement does not disclose specific licensing terms, pricing structure, or whether the arrangement is exclusive to Protege — the public framing is about access and rights-clearance, not deal economics.

For companies operating in adjacent data-licensing categories, this deal is a useful reference point on how a rights-holder-plus-platform-intermediary structure gets marketed publicly: emphasize scale (40 million videos), emphasize rights-clearance and transparency, and leave commercial terms out of the public announcement entirely. It's also one more confirmation that video specifically remains a premium, high-demand category — global video catalogs from established platforms are being actively courted for exactly the reason our own structured-interview data is valuable: verifiable rights and real diversity of source material are what AI labs are short on, not raw volume.

Key Points

• Partnership gives AI model builders access to ~40M videos from Dailymotion's multilingual catalog via Protege's platform
• Framed around a "licensed, transparent framework" with "rights-cleared access"
• Positioned to address scarcity of rights-protected, high-quality global video data
• Specific licensing terms, pricing, and exclusivity were not disclosed in the announcement