December 13th.OpenAI AI models for video generation Sora The source of the training data used has been confusing.All indications are that it may contain videos from Twitch live streams and gameplay walkthroughs. This discovery sparked a discussion about copyright and legal risks.
TechCrunch, a technology outlet, published a blog post on December 11th, reporting that after Sora opened up this past Monday, tests found that the generated videos used footage similar to games like Super Mario Bros, Call of Duty, Counter-Strike, and Teenage Mutant Ninja Turtles, and even featured the likenesses of Twitch anchors Auronplay and Pokimane. .
OpenAI has been tight-lipped about the source of Sora's training data, with legal experts pointing out that OpenAI could be at risk of copyright infringement if it uses gameplay footage for training without authorization.
In addition, experts believe that gameplay footage encompasses the copyrights of game content owned by the game developer and the copyrights of the producer of the footage, and may even involve the copyrights of user-generated content.