Recently, the copyright war in the AI industry has spread from the text domain to the video creation circle. According to TechCrunch, several well-known YouTube bloggers, including h3h3, MrShortGame Golf, and Golfholics, have formally filed a joint lawsuit against the social media giant Snapchat, accusing it of using their channel's video content without permission to train commercial AI models.

The lawsuit documents reveal the core dispute: Snapchat is accused of using a dataset called HD-VILA-100M to train its AI model. Although this dataset contains a massive amount of high-definition videos, its licensing agreement strictly limits use to academic and research purposes and explicitly prohibits any commercial activity. The plaintiffs claim that Snapchat violated the agreement by applying the model trained on this dataset to its commercial applications, such as Imagine Lens.

Currently, the plaintiff bloggers have applied to the court for statutory damages and have requested a permanent injunction against Snapchat to prevent the infringement from continuing. With the explosion of generative AI, such copyright lawsuits are increasing rapidly. According to the "Copyright Alliance," there have been more than 70 lawsuits against AI companies globally, although most are still in lengthy trials. However, this case may have significant implications for the industry in defining the commercial use of academic datasets.