Andersen v. Stability AI: What Artists’ Generative-AI Lawsuit Has—and Has Not—Established

The case survived in part, but procedural rulings are not final findings of infringement.

What happened—and what the record establishes

Visual artists sued Stability AI and other image-generation companies, alleging that copyrighted works were used in model development and raising copyright, trademark, and related claims. The court’s 2024 dismissal order allowed some theories to continue while dismissing or requiring amendment of others.

A claim surviving a motion to dismiss means the allegations were legally sufficient at that stage; it does not prove that training or outputs infringed. Later discovery has addressed model behavior, datasets, experts, and confidential technical information.

The litigation is important because it separates several questions often collapsed into one: copying during training, storage or representation in model weights, substantial similarity of outputs, removal of copyright-management information, and use of artists’ names as prompts or style references.

Questions the law must answer

• What protected works were copied and by whom?
• Are challenged outputs substantially similar to particular works?
• Can claims be proved separately for each defendant and system?

Practical takeaways

• Do not report a procedural ruling as a final victory.
• Copyright protects expression, not a general artistic style by itself.
• Creators should retain registrations, source files, publication dates, and suspected outputs.