Amazon sued over alleged YouTube scraping for Nova Reel training

Three YouTube creators have sued Amazon, alleging the company scraped millions of videos from the platform to train its Nova Reel video generation model. According to the complaint, Amazon used automated download tools, rotating IP addresses and virtual machines to bypass YouTube protections against large-scale extraction.

The plaintiffs are Ted Entertainment, which operates h3h3 Productions and H3 Podcast Highlights, along with golf creators MrShortGame Golf and Golfholics. They argue that Amazon violated both YouTube's terms and the DMCA's anti-circumvention rules. They are seeking damages as well as an injunction to stop further use of the material. Amazon has declined to comment on active litigation.

The case lands in the middle of a widening fight over AI training data. Many AI companies have relied on the idea that publicly accessible content can still be used for model training. This lawsuit goes after the collection method itself, not just the downstream model output. If the court finds that Amazon actively bypassed digital safeguards to gather video data at scale, the fallout could reach far beyond Nova Reel.

For CIOs and AI leaders, the takeaway is straightforward: data sourcing is turning into a governance issue, not just a technical one. As models become more agentic and more multimodal, the risk around where training data came from only gets bigger.