Warner Bros. has filed a high-stakes lawsuit against AI image startup Midjourney, accusing the company of enabling users to generate unlicensed images and videos featuring Superman, Batman, Bugs Bunny, and other studio-owned characters. The complaint, first noted by Reuters, squarely targets what the studio calls a systemic disregard for copyright controls and a deliberate easing of safeguards as the platform’s capabilities expanded.
What Warner Bros. Alleges
The studio argues Midjourney not only hosts a powerful generative system capable of outputting highly recognizable, protected characters, but also made policy choices that invited infringement. According to the complaint, Midjourney previously constrained prompts tied to known IP but later relaxed those rules, allowing subscribers to produce material that tracks closely with DC and Looney Tunes properties.
Warner Bros. seeks monetary damages, the return of profits allegedly tied to unauthorized outputs, and a court order forcing tighter restrictions. While the centerpiece is copyright, the case could also implicate trademark and unfair competition theories—particularly if outputs cause consumer confusion or dilute famous marks linked to the studio’s franchises.
Midjourney’s Expected Defense
Midjourney has contended in prior disputes that training on publicly available images can qualify as fair use under U.S. law. Its argument typically distinguishes between ingesting data to learn patterns and distributing the original works. The company also tends to frame outputs as user-directed, emphasizing that it does not host or sell the underlying characters themselves but generates new images based on text prompts.
However, the fair use debate looks different when the outputs reproduce or closely evoke protected characters. U.S. copyright law treats well-delineated characters as protectable expression, and studios often pair those rights with trademark protection to police brand associations. The legal question is whether an AI-generated image of Superman is a transformative parody or a derivative work that competes with licensed uses—a boundary courts have not fully settled for generative systems.
Parallel Battles Shaping the Landscape
The case follows a suit filed by Walt Disney and Universal that targets Midjourney for similar uses involving characters such as Darth Vader, Bart Simpson, and Shrek. Those filings highlight a broader rights-holder push to curtail unlicensed character generation, signaling a coordinated front across major media companies.
Beyond Midjourney, courts are already grappling with adjacent questions. Getty Images has sued Stability AI over training on watermarked photographs, and a group of artists in Andersen v. Stability AI has challenged training and outputs as infringing derivative works. Meanwhile, a landmark Supreme Court decision in Warhol v. Goldsmith narrowed how “transformative” purpose is weighed in fair use, underscoring the importance of market substitution—an analysis that may loom large when fan-favorite characters are recreated without a license. Separately, the U.S. Copyright Office has reiterated that fully AI-generated works are not eligible for copyright, while continuing a policy study on training data transparency and compensation frameworks.
Why Characters Are a Different Kind of IP
Iconic characters are less like generic styles and more like branded assets. Superman’s emblem, Batman’s cowl, and Bugs Bunny’s likeness are tightly controlled across films, TV, toys, and games. Licensing International estimates global retail sales of licensed merchandise at well over $350 billion annually, and studios invest heavily to protect the integrity and value of those brands.
Generative tools can produce faithful replicas or edgy reinterpretations at scale, including content that would never clear a studio’s legal or brand review. From the rights-holder perspective, that undermines quality control, confuses consumers, and chips away at premium licensing deals. For the AI developer, over-aggressive filtering risks throttling user creativity and growth, while under-filtering invites litigation and potential injunctions.
What’s at Stake for AI Platforms
If Warner Bros. secures an injunction, Midjourney could face immediate pressure to implement granular filters around character names, visual motifs, and lookalike outputs. Some peers have pursued licensing strategies—such as training on stock libraries or approved datasets—paired with content credentials and provenance signals. Industry groups are backing technical standards like C2PA to label AI outputs, and policy discussions at the U.S. Copyright Office have explored disclosure obligations and opt-out mechanisms for training data.
Discovery could force unprecedented detail about Midjourney’s training corpus, safety tooling, and commercialization. That, in turn, may influence how other model builders weigh the trade-offs between open-ended generation and rights-compliant guardrails. A settlement with tiered licensing for major franchises is also conceivable, particularly if the court views character-level generation as high-risk without permission.
Bottom line: this lawsuit is a referendum on how far generative AI can go with cultural icons that studios regard as crown jewels. However the court resolves it—through injunction, damages, settlement, or a path to licensed generation—the outcome will shape the playbook for IP compliance across the most valuable characters in entertainment.