Openai has released Sora 2, the latest version of its video generation model, but the tool has attracted criticism for creating content featuring already copyrighted characters.
Released on September 30th, SORA 2 allows users to generate shorter forms of animation clips containing familiar figures such as Pikachu, Goku, and Spongebob Squarepants.
According to 404 media, this model is used to create videos of characters in unusual and inappropriate scenarios, such as ASMR content, boxing matches, and in some cases Nazi versions of SpongeBob.
By default, the system is trained with copyrighted materials and leaves it to the right holder responsible for requesting deletion rather than requesting prior consent.
This approach put Openai at the centre of a broader discussion about generative AI and intellectual property. While Disney, NBC and Warner Bros. took action against Midjourney, similar lawsuits were already underway, and artists who contributed to the magic: The Gathering also complained of misuse of artwork in their training datasets.
Beyond copyrighted characters, SORA 2 has also been reported to generate deepfakes for public figures and ordinary users despite claims of protection.
Some users claim they saw manipulation content that includes real celebrities, including instances of the blackface version of Scarlett Johansson. Openai previously faced a controversy over the misuse of a voice similar to Johansson on its ChatGpt product.
Critics argue that tools like SORA 2 may encourage studios to reduce their reliance on human animators in favor of cheaper AI output. For many artists, this is not only a legal issue, but also a challenge to the creative integrity of their work.
Pokémon Company, Supergiant Games and Toei Animation have been contacted for comments on the character's appearance in the Sora 2 video. Given the history of legal challenges in this field, analysts suggest that further conflicts could be on the horizon.
