Part I: Disney v. Midjourney
On June 11, 2025, Disney and several affiliated production companies filed a federal lawsuit against Midjourney, Inc., a leading artificial intelligence (AI) image-generation platform. The suit alleges “calculated and willful” copyright infringement—marking the first major clash between Hollywood and a generative AI provider.
According to the complaint, Midjourney generates and distributes unauthorized images of Disney characters and other intellectual property (IP) owned by Disney and its affiliates. The complaint includes side-by-side comparisons of Disney’s original works and AI-generated replicas, offered as direct evidence of copyright infringement.
The plaintiffs allege that:
- Midjourney’s model was trained using copyrighted content scraped from the internet without proper consent.
- The platform enables users to produce derivative works that infringe on protected IP.
- Midjourney has profited significantly, earning over $300 million in 2024, by allowing and monetizing this allegedly infringing activity.
Disney’s chief legal officer, Horacio Gutierrez, emphasized the company’s position: “Piracy is piracy, and the fact that it’s done by an AI company does not make it any less infringing.”
Although Midjourney has not formally responded in court, its founder, David Holz, has previously acknowledged the inclusion of web-scraped images in the platform’s training dataset. He noted the difficulty of verifying the provenance of large-scale image collections, stating, “There isn’t really a way to get a hundred million images and know where they’re coming from.”
Disney’s lawsuit seeks monetary damages and an injunction that could halt Midjourney’s operations—just as the company prepares to launch a commercial video-generation product.
Part II: Bartz v. Anthropic
Now, a federal court decision issued just two weeks after Disney sued Midjourney (on June 23, 2025) may complicate Disney’s claims. In Bartz v. Anthropic, the court held that training a generative AI model using copyrighted books—where the books were used to develop statistical relationships for generating new, original outputs—constituted a “quintessentially transformative” fair use under Section 107 of the Copyright Act. The court emphasized that the AI outputs did not reproduce or distribute infringing content and that the training process, while involving multiple copies of the works, served a fundamentally different purpose than the original works.
Although the Anthropic decision did not excuse unauthorized acquisition of pirated works for building a general-purpose library, it strongly favored fair use where the copyrighted content was used specifically for model training. This precedent may bolster Midjourney’s defense if it can demonstrate that its model outputs do not recreate protected content and that any training use of copyrighted material was similarly transformative.
Part III: Why This Matters & Recommendations
Whether building or deploying generative AI, companies should closely monitor the Disney v. Midjourney litigation and evolving case law, which may reshape risk assumptions around copyright, licensing, and liability.
In the interim, companies developing or deploying generative AI systems should take proactive steps to mitigate risk in light of emerging case law. Consider the following risk-mitigation strategies:
- Review sourcing practices for training data: Ensure datasets are lawfully acquired—the Anthropic decision reaffirms that using pirated or improperly sourced content remains a clear infringement, even if the end use is transformative.
- Assess whether training use is transformative and isolated: Where training uses are clearly aimed at enabling models to generate new, non-infringing outputs (and outputs are filtered appropriately), courts may find such uses to be fair. But storing or reusing copyrighted works for general purposes or future uses remains problematic.
- Document and limit retention of source material: Maintain records of what data is used for model training and avoid indefinite retention of content not actively in use—especially if it was not acquired through a license or public domain.
- Establish output filtering and moderation protocols: Implement controls to prevent the model from generating infringing or derivative outputs, a factor the Anthropic court noted in favor of fair use.
- Clearly allocate responsibility in contracts: Assign responsibility for model behavior, content moderation, and indemnification in vendor and partner agreements.
To evaluate the risks and compliance obligations that generative AI may create for your business, contact the authors or any attorney with Frost Brown Todd’s Technology Transactions, Artificial Intelligence (AI), Copyright, or Intellectual Property teams to discuss practical strategies and legal safeguards.