In their recent lawsuit against AI image generator Midjourney, Disney and Universal are not only seeking damages and declaratory relief as is typical in the numerous copyright infringement cases pending against generative AI providers. They are also asking the court to require Midjourney to adopt technical safeguards that would prevent users from generating infringing content in the future. This type of remedy looks ahead rather than backward and reflects a growing effort by rightsholders to impose structural obligations on AI platforms that are enabling unauthorized use of copyrighted material.

The complaint asks that Midjourney implement filtering tools to block generation of images based on the plaintiffs' well-known characters, such as Darth Vader, Iron Man, and the Disney princesses. Disney and Universal point out that other AI providers have adopted similar tools and accuse Midjourney of choosing not to do so in order to unfairly maximize its profits. The lawsuit emphasizes that Midjourney’s refusal to add these filters is not an oversight but a deliberate business decision, and one that has allowed the service to distribute infringing material at scale.

This request stands apart from what many other plaintiffs have focused on in prior lawsuits against AI companies. In most cases, the attention has been on the training phase, where copyright owners have alleged that the AI models were trained using protected works without permission and have asked for damages and/or injunctions based on that use. Disney and Universal's additional requested remedy, however, shifts the spotlight to the platform’s current operation. Their claims center on what users are able to generate today and the lack of controls that would prevent those outputs from violating copyright law.

Midjourney might argue that it should not be liable where its service, like the VCR at issue in Sony Corp. of America v. Universal City Studios, Inc., is a tool capable of both infringing and noninfringing uses. In Sony, the Supreme Court held that the manufacturer of a technology cannot be held liable for contributory copyright infringement if the product is widely used for legitimate purposes and only incidentally enables infringing uses. Midjourney may point to the many lawful and creative applications of its image generator as evidence that the service has substantial noninfringing uses, such as generating original artwork or concept designs not tied to existing intellectual property.

Disney and Universal, however, may counter that unlike the passive copying function of a VCR, Midjourney’s system actively interprets user prompts and generates outputs that closely resemble protected characters. This makes the connection to infringement more direct and foreseeable. In addition, Midjourney does not merely provide a medium on which infringement might occur, but is instead the source of the allegedly infringing material itself.

If this and any other defenses are unsuccessful, the remedy sought could matter even more if courts ultimately rule that using copyrighted works to train AI models is fair use. In that event, developers might not be liable in connection with the use of copyrighted materials to build models. But they could still be required to take steps to stop their tools from producing and distributing infringing material. The outcome of this case could thus set a new baseline for what is expected of AI companies, even if the law allows broad freedom during training.