AI Policy Wiki
Dashboard

Andersen v. Stability AI

medium confidence · updated 2026-08-12

The first US generative-AI copyright class action (N.D. Cal., filed January 13, 2023) — visual artists against Stability AI, Midjourney, Runway AI and DeviantArt over image-dataset training and output-infringement theories. Survived Rule 12, still in discovery as of July 2026.

Andersen v. Stability AI Ltd., No. 3:23-cv-00201 (N.D. Cal.), is a putative class action brought by visual artists against Stability AI, Midjourney, Runway AI and DeviantArt over the use of copyrighted images to train text-to-image generative models. Filed on January 13, 2023 by Sarah Andersen, Kelly McKernan and Karla Ortiz through the Joseph Saveri Law Firm, it is the earliest of the generative-AI copyright suits and preceded the December 2023 New York Times complaint by nearly eleven months (Source: courtlistener.com).

Infobox

FieldValue
Case number3:23-cv-00201 (N.D. Cal.)
FiledJanuary 13, 2023
Original plaintiffsSarah Andersen, Kelly McKernan, Karla Ortiz
Plaintiffs on the operative complaintAndersen, Ortiz, McKernan plus Gerald Brom, Adam Ellis, Julia Kaye, Gregory Manchess, Grzegorz Rutkowski, Jingna Zhang
DefendantsStability AI Ltd., Stability AI Inc., Midjourney Inc., Runway AI Inc., DeviantArt Inc.
Plaintiffs' counselJoseph Saveri Law Firm
Magistrate judge (discovery)Lisa J. Cisneros
StatusActive; in discovery
Docket entries691 as of July 31, 2026

Background

The suit concerns Stable Diffusion and models derived from or built on it. The plaintiffs allege that the models were trained on billions of images scraped from the web, including their own copyrighted works, without licence or compensation, and that the resulting systems can produce outputs in the style of individual named artists. DeviantArt is named for deploying a Stable Diffusion–derived product to its user base; Midjourney and Runway AI are named as builders of image-generation products the plaintiffs allege incorporate the same training.

Two theories run through the case and recur across the wider docket surveyed at AI Copyright Litigation — Analysis. The ingestion theory treats the act of copying works into a training corpus as infringement in itself. The output theory treats particular generated images as infringing derivatives or reproductions. The case also raised Lanham Act trade-dress claims tied to artists' names being usable as style prompts, and DMCA § 1202 claims over removal of copyright-management information.

Procedural history

The original complaint was filed January 13, 2023 with a jury demand and seventeen exhibits. The court substantially narrowed the case at the pleading stage, dismissing most claims against the non-Stability defendants with leave to amend while allowing the direct-infringement claim against Stability to proceed; the plaintiffs amended, adding six further named artists and re-pleading the induced-infringement and Lanham Act theories. The narrowed case survived subsequent challenges and moved into discovery (AI Copyright).

As of July 31, 2026 the docket ran to 691 entries, with activity in the final week of July 2026 consisting of discovery letter briefs, sealing motions and a transcript of proceedings — a posture consistent with contested discovery rather than dispositive briefing (Source: courtlistener.com).

Current status

The case is active and has not produced a merits ruling on fair use. It is the longest-running of the generative-AI copyright actions, and the only one of the three January–June 2023 filings still litigating in its original court: the parallel Getty Images US case was voluntarily dismissed in August 2025, and Tremblay v. OpenAI was transferred into a multidistrict litigation in April 2025.

Relationships