US judge pumps the brakes on artists' AI copyright lawsuit

Artists thought generating images in their signature style was a slam-dunk copyright case against AI giants. A federal judge just reminded them that copyright law does not work on vibes.

U.S. District Judge William Orrick signaled that he is inclined to dismiss the majority of a high-profile copyright lawsuit brought by a group of artists against Stability AI, Midjourney, and DeviantArt. However, he is giving the plaintiffs a chance to file an amended complaint.

The suit, initially filed in January by artists Sarah Andersen, Kelly McKernan, and Karla Ortiz, alleged that Stability AI scraped billions of images from the web without consent to train its Stable Diffusion model, enabling users to generate works copying their styles.

A lack of evidence and the style problem

Judge Orrick pointed out glaring flaws in the artists' legal argument. First, the court noted that Stable Diffusion was trained on roughly five billion compressed images. Without concrete factual evidence, the judge found it implausible to simply assume the plaintiffs' specific works were part of that training run.

Orrick emphasized that because the plaintiffs have access to Stability AI's relevant source code, they need to bring real, verifiable facts showing infringement rather than broad assumptions.

The biggest hit came on the output side. The artists argued that AI-generated images prompted with their names infringed on their rights. Orrick shut that down, stating there is currently no substantial similarity between the artists' original creations and the AI-generated outputs.

The claims against Midjourney and DeviantArt were also deemed murky. Orrick said it remains unclear whether the artists are accusing those platforms of direct copyright infringement for training their own systems, or vicarious liability for implementing Stability AI's technology.

The only claim likely to survive this initial hurdle belongs to illustrator Sarah Andersen, specifically because she holds formal copyright registrations for the works she claims Stability AI directly infringed.

Why it matters

For artists and creators, this ruling is a reality check. You cannot simply point to an AI mimicking your artistic style and cry copyright infringement, because copyright protects specific expressions, not generic styles. Unless creators have formally registered copyrights and can demonstrate that an output is substantially similar to an existing work, taking on AI companies in court will be an uphill battle.

For the AI industry, this provides significant breathing room. Tech giants like OpenAI, Microsoft, and Meta are facing a wave of similar training-data lawsuits. By maintaining a strict traditional standard for substantial similarity, the court signaled it won't easily bend existing law to penalize AI models for stylistic imitation.

We think the judge made the correct legal call under current statutes, but it highlights a massive structural imbalance. Expecting independent artists to pinpoint their individual art within a dataset of five billion compressed images is an unrealistic burden of proof. AI companies may have won this round on procedure, but the core debate over data consent is far from settled.