Whether you call it the Italian plumber problem or the Snoopy problem may depend on which character dominated your own childhood imagination. In the latest AI lawsuit, brought by Disney and Comcast against image generator Midjourney, it’s a Minions problem, a Vader problem, a Spider-Man problem, and even a little bit of an Elsa-from-Frozen problem. But what’s the problem? In a word: characters.

In a few words—Matt Sag’s words—the problem is that “the more abstractly a copyrighted work is protected, the more likely it is that a generative AI model will “copy” it.” AI is probabilistic—its outputs are the result of averaging across many thousands of training inputs, so an AI tool’s outputs are extremely unlikely to be copies of any specific image or other work from its training data. But characters like Snoopy are protected abstractly—an image can be an infringing “copy” of a character even if it is not a literal copy of any previous image. 

Giving users the ability to create new images with old characters seems to be what really got Midjourney in trouble, and press coverage of the lawsuit shows the effectiveness of Disney’s complaint as an advocacy tool. Multiple outlets reprinted their exhibits showing side-by-sides of Disney originals and Midjourney outputs of a laundry list of beloved characters. As Copyright Lately’s Aaron Moss put it, the argument is basically, “Just look what this thing produces.” Q.E.D., right?

As a kid who grew up reading Mad Magazine parodies and drawing Wolverine all over my school notebooks, I’m not so sure. Do we really want to let Disney and Kabletown decide who gets to make new images that incorporate recognizable character imagery? If the gravamen of this complaint is that Midjourney dared to create a tool capable of making new images with old characters, I hope the court will remember Alfred E. Neuman and consider that general purpose creative tools like Midjourney are capable of many lawful uses, including unlicensed uses of copyrighted characters.