Generative AI is still plaguing every creative industry you care about, and despite the obvious copyright infringement issues and related legal cases surrounding it, companies like OpenAI are still using copyrighted works of art to train their models. OpenAI's video generation model Sora has launched an updated "Sora 2" version on September 30.

The application will use copyrighted materials for self-training by default, and requires copyright holders to actively apply to withdraw in order to prevent their materials from being used. As a result, tons of AI-generated content featuring popular characters you know and love is leaking onto the internet.

OpenAI video generation model Sora 2 is online and a large number of popular character videos appear

404 Media obtained videos of Pikachu and a Nazi-like Spongebob doing everything from ASMR to boxing matches. By default, Sora 2 can generate animated videos of just about any copyrighted material you feed in — because generative AI companies seem to think that the rules that usually govern the use of copyrighted intellectual property don't apply to them. So now, companies like Nickelodeon and Nintendo have to contact OpenAI and ask it to stop using their characters and icons, instead of OpenAI asking for permission. This strategy is bold considering that Disney, NBC and Warner Bros. are all suing Midjourney for using characters from their intellectual property. Artists involved in the art creation of the Magic card game also sued Midjourney because the company grabbed their artwork to train generative AI models.

In addition to this low-quality content containing copyrighted characters, Sora 2 appears to allow users to generate deepfake content of themselves or public figures. The new app features a TikTok-like feed where users can scroll infinitely through generated videos; and if you scroll long enough, you might see deepfake videos of real people — although OpenAI claims to have safeguards in place to protect people from this. Some people have even reported seeing outrageous content, such as actress Scarlett Johansson performing in blackface in the musical "Hamilton." It’s worth noting that OpenAI had a legal dispute with the actor last year when the company used a voice in its ChatGPTAI chatbot that was very similar to the voice of Scarlett Johansson’s AI character in the movie “Her.”

The whole thing seems destined to spark a slew of legal battles, as well as frustration for some artists and animators, as it emerges that animation companies may increasingly turn to generative AI to cut costs at the expense of human touch.

Even the smoother, more action-intensive scenes generated by Sora still have obvious traces of AI generation, such as unnatural screen transitions and glitchy effects in the animation. But for those looking to use it to cut costs, quality may not be a priority. Those big guys who insist that AI is the future say so only because they are the ones who can benefit from it, while the artists who create and the people who appreciate these works have become the losers. Maybe some intellectual property lawyers will step in and try to stop this, but sometimes it feels like every time a tool for generating AI-bad content is outlawed, another one takes its place.