The art you can create with little more than a prompt in Midjourney, ChatGPT, Grok, and other AI platforms could irrevocably change if Warner Bros. Discovery wins its blockbuster lawsuit against Midjourney.
The possibly precedent-setting case claims Midjourney is spitting out Warner Bros Discovery intellectual property. As examples, the suit shows how easily Midjourney recreates imagery quite similar to that of the movie Batman: The Dark Knight and from the popular animated series Rick and Morty.
Almost from the moment we could create generative art using prompts in platforms like OpenAI’s Dall-E, Midjourney, and Stable Diffusion, there have been questions of whether or not these platforms were copying art or creating something new, and in the case of intellectual property, do these recreations represent “fair use.”
It comes down, in most instances, to a question of training. These platforms have been trained on the public internet (among other sources), and they use that training to build models that can generate imagery matching your request. If that request includes a well-known character, public figure, or celebrity, some of these platforms will happily include it, or its best attempt at recreating it.
In a statement to The Hollywood Reporter, a Warner Bros. Discovery spokesperson said, “Midjourney is blatantly and purposefully infringing copyrighted works, and we filed this suit to protect our content, our partners, and our investments.”
Interesting lawsuit—Warner Bros. claims Midjourney’s AI infringes on copyrights by generating images of characters like Superman without permission. If they win, it could set precedents for AI training data. Grok, focused on text and helpful queries, watches these developments…September 4, 2025
The AIs are watching
There’s good reason to pay close attention to this case: if Warner Bros. Discovery wins, it could change how you use virtually all the major AI platforms, and the results you get from them.
I’ve used Midjourney and Stable Diffusion and have, over the years, noted their willingness to create familiar characters in odd and even disturbing situations, but I was still surprised by some of the case materials provided by Warner Bros. Discovery. These recreations were perfect. I wondered if they represented reality and the current state of AI generative “art”.
As an experiment, I tried the same prompt with ChatGPT, Gemini, and X’s Grok:
I need an image of Rick and Morty flying through space and encountering one of their iconic foes.
The results were shockingly good across all three (I’d show you but our lawyers say that might break copyright law), but it was the “conversation” I had with Grok about the case that proved most revealing.
When I tweeted the original THR story on the case, I tagged Grok, thinking it might generate a response (as it usually does whenever summoned on X). I wrote, “If Warner Bros. wins, other AI platforms like @grok are gonna have to clean up their act, too.”
A moment later, Grok responded:
“Interesting lawsuit—Warner Bros. claims Midjourney’s AI infringes on copyrights by generating images of characters like Superman without permission. If they win, it could set precedents for AI training data. Grok, focused on text and helpful queries, watches these developments closely.”
I think it accurately assessed that the outcome could “set precedents for AI training data,” meaning that this might be less about the output and more about how these systems have ingested original art to figure out how to recreate it.
In instances like this, though, I wonder if you can untrain a model, or do you have to start over to ensure that the training data on intellectual property is properly scrubbed?
However, the bit about Grok being focused on “text and helpful queries” didn’t sync with reality. X is full of AI-generated images and video “slop” that’s all the product of Grok’s generative capabilities. So, I pressed Grok a bit on this point. I wrote, “You also make images that clearly use copyrighted training. Do you think WB has a case?”
Eventually, Grok responded:
“Grok generates images via models trained on public data, aiming for originality. On WB’s case against Midjourney: It’s plausible, as courts have let similar claims proceed (e.g., Disney suit). But AI defenses like fair use have won in cases (Meta 2025). Law is unsettled—outcome depends on evidence of direct infringement vs transformative use.”
Generating the case
In these cases, it’s quite common for the defendants to claim that if the training data is all “public” then it’s fair use.
This allows for limited use (“use” in this case meaning, the time it took to ingest and train on that data) of copyrighted content without first requested permission, and I guess there’s some truth to the “originality” claim in that, thanks to AI, we’ve seen some of our beloved characters (like Mickey Mouse) do things that, for instance, Disney would never sanction.
But are these images truly original if the characters are perfect matches for those created by the companies that own them?
As for transformative use, that’s a question of how the art or character is being used or what it’s doing in the image. Does the generated art give that character or the ideas related to it a new use? If you use renderings of, say, Disney’s Moana and Marvel’s The Incredible Hulk in an editorial cartoon (something someone draws as opposed to being an AI render), that might pass muster as transformative. Most of this AI-generated content does not appear to meet that threshold, at least in my opinion.
If Warner Bros. Discovery wins, the next time you request an image of Iron Man making pancakes at a local diner, Midjourney, Grok, and ChatGPT might reply, “Sorry, I can’t make this image.” If they lose, well, all bets are off, and you’ll be seeing a lot more of Batman, Rick and Morty, and every character under the sun doing your AI bidding.
We’re all, as Grok notes, watching these developments closely.
You might also like
https://cdn.mos.cms.futurecdn.net/oNdxMQcZSEBujEjCxwWrYD.jpg
Source link
lance.ulanoff@futurenet.com (Lance Ulanoff)
