On September 4, 2025, Warner Bros. Discovery filed a copyright infringement lawsuit against Midjourney Inc. in the U.S. District Court for the Central District of California. The complaint was filed on behalf of Warner Bros., DC Comics, Turner Entertainment Co., Hanna-Barbera Productions Inc., and the Cartoon Network. The characters at issue include Superman, Batman, Wonder Woman, Scooby-Doo, Bugs Bunny, and a long list of other properties spanning DC Comics, Looney Tunes, and Hanna-Barbera.

The complaint's core allegation: Midjourney trained its image generation model on Warner Bros.' copyrighted works without a license, and its tool produces images that are recognizable copies of those works. Warner Bros. called the infringement "systematic" and "willful," and accused Midjourney of believing it is "above the law."

Hollywood's Third Strike

Warner Bros. was not the first major studio to sue Midjourney. That distinction belongs to Disney and Universal, which filed a joint lawsuit against Midjourney on June 11, 2025, in the same court. The Disney/Universal complaint called Midjourney a "bottomless pit of plagiarism" that generates "endless unauthorized copies" of their work. It included side-by-side comparisons of original Disney and Universal characters next to Midjourney outputs - and those outputs were not generated by attorneys crafting careful prompts to provoke infringement. They were generated by ordinary Midjourney users doing ordinary things, like asking what Shrek would look like as a 1950s greaser.

Warner Bros.' complaint followed a similar structure. By September 2025, three of Hollywood's "Big Five" studios had sued Midjourney, with only Sony and Paramount remaining on the sidelines. That all three chose to target Midjourney specifically - rather than other AI image generators like Stability AI or DALL-E - said something about Midjourney's particular approach to copyright compliance, or lack of it.

The Takedown Problem

A central allegation in the Warner Bros. complaint was that Midjourney ignored takedown requests. The studios said they had asked Midjourney to stop generating images of their copyrighted characters and to implement technological measures to prevent it. Midjourney, according to the complaint, did neither.

This matters because AI companies facing copyright lawsuits often argue they can't control what users prompt their tools to create. The standard defense is some variation of "we're a platform, users are the ones generating the images, we can't police every output." But that defense weakens considerably when the platform has been notified of specific infringements and asked to take specific steps to prevent them and has declined to do so.

In traditional DMCA takedown scenarios, platforms have a well-established process: receive a notice, remove the content, notify the uploader. AI image generators don't fit neatly into that framework because the "content" isn't a static file sitting on a server waiting to be taken down. The infringing material is, in a sense, baked into the model's weights. A user doesn't need to upload Batman to get Batman - they just type "Batman" and the model produces something that looks like Batman, because the model was trained on images of Batman that belonged to DC Comics.

Warner Bros.' position was that Midjourney had the ability to filter prompts and reject outputs that matched known copyrighted characters, and simply chose not to. Other AI image generators had implemented at least some prompt-level filtering for copyrighted content. Midjourney's failure to do so, combined with its non-response to takedown requests, formed the basis for the "willful" infringement allegation, which carries significantly higher potential damages under copyright law.

What Warner Bros. Wanted

The lawsuit sought statutory damages of up to $150,000 per infringed work, plus an injunction that would force Midjourney to purge its models of the copyrighted training data. Warner Bros. also requested a jury trial.

The statutory damages request was the financial sledgehammer. Under U.S. copyright law, willful infringement carries statutory damages of up to $150,000 per work. When your intellectual property portfolio includes Batman, Superman, Wonder Woman, the entire Looney Tunes roster, the Hanna-Barbera catalog, and everything else Warner Bros. controls, the number of individual copyrighted works involved is enormous. Multiply $150,000 by hundreds or thousands of individual copyrighted works and the theoretical damages figure quickly enters the billions.

The injunction request was the operational sledgehammer. Forcing Midjourney to purge copyrighted training data from its models would effectively require retraining from scratch without the offending material - or shutting down the parts of the model that produce the infringing outputs. Neither option is trivial.

The Fair Use Question

Midjourney's expected defense, based on its response to the Disney/Universal lawsuit, was fair use and the argument that it cannot control what users create. AI companies facing copyright litigation have generally relied on the fair use doctrine, which allows limited use of copyrighted material under certain conditions. One of the key factors courts examine is whether the new use is "transformative" - whether it adds a new meaning or message to the original work.

For text-based AI models, the fair use argument had some momentum. In the Bartz v. Anthropic case, Judge William Alsup ruled that using copyrighted books to train a text-generating AI was "quintessentially transformative" because the AI's outputs were fundamentally different from the training data - the AI produced new text, not reproductions of the books.

Image generation presented a different problem for the fair use defense. When a user asks Midjourney to generate "Batman on a rooftop in Gotham City," the output is not a transformation of Batman into something wholly new. It's a picture of Batman on a rooftop in Gotham City. The character design, the costume, the setting - these come from DC Comics' decades of creative work, and the AI output reproduces their essential elements.

IP attorney Chad Hummel, commenting on the Disney/Universal case, called the complaints "extremely significant" and said the compilation of evidence suggested "the output is not sufficiently transformative." The core distinction between text AI and image AI cases may come down to the recognizability of the output. A language model trained on Harry Potter doesn't produce Harry Potter when you ask it a question. An image model trained on Batman produces Batman when you ask for Batman.

Midjourney's Position

Midjourney, founded by David Holz, had been comparatively muted in its public response to the lawsuits. The company operated differently from other major AI firms. It distributed its image generator through Discord, running as a bot that responded to text prompts in chat channels. It had no traditional corporate website with a legal team issuing public statements. It had no press office releasing carefully worded responses to litigation.

Midjourney's approach to training data had been blunt. Like many AI image generation companies, it trained its models by scraping the internet to build large datasets of images. It did not seek individual licenses for the copyrighted works in those datasets. The legal theory underpinning this approach - that training constitutes fair use - was untested in court when Midjourney built its models. By September 2025, it was being tested aggressively.

The McKool Smith law firm, tracking the case, noted that Midjourney's answer would likely follow the same pattern as its response to Disney/Universal: claiming fair use and arguing that it does not control what users create. The legal community expected no surprises in the defense strategy. The question was whether that strategy would hold up when the evidence included output after output of recognizable copyrighted characters generated by ordinary users through ordinary prompts.

The Industry Alignment

The Recording Industry Association of America endorsed the Disney/Universal lawsuit as a "critical stand for human creativity and responsible innovation." The music industry had been fighting its own parallel battles against AI companies generating synthetic music, and the Hollywood studios' lawsuits gave the broader creative industry a unified front against unlicensed AI training.

Within Hollywood, the significance was in the coalition. Disney, Universal, and Warner Bros. are direct competitors. They fight over box office weekends, streaming subscribers, and merchandising shelf space. For three of the five largest entertainment conglomerates to align on a single legal target reflected something broader than any individual company's grievance - it reflected an industry consensus that Midjourney's approach to copyright was unacceptable.

The lawsuit also landed in a legal environment that was shifting fast. The Anthropic fair use ruling gave AI companies some cover for text-based training, but the image generation cases raised different questions. If the Midjourney cases reached trial, the courts would need to decide whether producing recognizable copies of copyrighted characters is "transformative" in the same way that training a text model on books is transformative. The answer to that question could reshape how every AI image generator operates.

Warner Bros. brought Batman to court. Whether the court sides with the studio or the algorithm will say a lot about what copyright means in an era when a machine can draw anything you describe, using creative work it never paid for.