More Studios Seen Likely to Follow Disney/Universal's AI Litigation

It's one of a host of pending suits that content creators have brought against generative AI platforms. Anthony Justice, a truck driver and independent country artist, sued Udio.com and Suno last week in federal courts in New York and Boston (dockets 1:25-cv-5026 and 1:25-cv-11739, respectively) over the use of copyrighted music to train their AI music generators. The suits followed similar ones that major music labels brought in 2024 in the same courts (docket 1:24-cv-04777 and 1:24-cv-11611, respectively).

Mike Schuster, a University of Georgia business associate professor, said Disney and Universal have a stronger case against the Midjourney outputs -- the images it generates -- than over the use of IP as training data. Also, Midjourney could fix the output issues using filtering software, he said.

The questions around Midjourney’s inputs -- its use of copyrighted material such as images in its training data -- are more nuanced and involve less-tested legal doctrine, Schuster said. There will be defense arguments about fair use exceptions to copyright, which allow unlicensed copying in limited circumstances, he said. A big question around a fair use exception is whether the copyrighted work is being transformed for a new purpose, Schuster noted, and in the Disney/Universal/Midjourney case, there are strong arguments either way. It’s still unclear how courts will proceed on this novel series of questions, he added.

Disney and Universal chose not to sue OpenAI or another AI behemoth because they have tremendous resources and legal savvy, said Mark Bartholomew, an associate law professor at the University at Buffalo, so a smaller company like Midjourney makes more sense as a legal target.

Shubha Ghosh, director of Syracuse University’s Intellectual Property Law Institute, said Sony has been an aggressive plaintiff on copyright issues, and it wouldn't be surprising if the company became active in AI-related litigation. Sony, the Motion Picture Association and the AI Association didn't comment.

However, Bartholomew said, targeting Midjourney presents special challenges. Unlike The New York Times' litigation against OpenAI and Microsoft, which concerns their use of Times content to train their large language models, Midjourney might argue that it's a passive conduit and that others are committing the actual infringement. Still, plaintiffs could have a strong case if they can show that Midjourney isn't taking proper precautions, since it's well-established that a business acting to help a user infringe copyrights is just as liable, he noted.

Ghosh told us that when it comes to using copyrighted work as an input, "intermediate copying" is less egregious than full-scale copying, such as piracy. He pointed to the 2nd U.S. Circuit Court of Appeals' Google Books decision (see 1510160063) as a key example of intermediate copying, with the court allowing Google to scan and make digital copies of books for its searchable database. But with Midjourney, the copies that it's making as input for its training data are resulting in outputs that are largely replicas and arguably not transformative, Ghosh said.

Courts deciding that AI companies must license all training data or avoid reproducing copyrighted elements would mean major challenges for Meta and its forthcoming AI-based ad-creation platform, public relations consultant Chris Gee blogged last week. The Disney/Universal case could result in standards for how much AI-generated content must differ from the original to avoid copyright problems, he blogged. Meta must show filtering that can stop AI from generating ads that too closely resemble copyrighted characters or visuals and clearly demonstrate that outputs were meaningfully transformed, he added.