Andersen v. Stability AI
Filed in the Northern District of California, the lawsuit alleged copyright infringement on a scale the existing legal framework was not designed to address. Users of Stable Diffusion could prompt the system to produce images in the style of Sarah Andersen, approximating her visual signature closely enough to compete with her for commissions. The three named plaintiffs represented a potential class of millions of artists whose work had been scraped from portfolio sites and social media platforms without consent, notification, or compensation. The lawsuit was not filed under cover of darkness by anonymous rioters; it was filed publicly, through the formal legal system. But its structural function was identical to the direct actions Thompson documented: the assertion of a claim — our labor has been taken without consent and used to enrich others — through the only mechanism available when the customary norms governing the use of that labor had been violated.
In the AI Story
The lawsuit was initially met with significant legal skepticism. In October 2023, the court dismissed most of the claims, ruling that plaintiffs had not adequately specified which of their works appeared in training data. In August 2024, the court allowed direct infringement, DMCA, and vicarious infringement claims to proceed against Stability AI, Midjourney, DeviantArt, and Runway. The partial reinstatement demonstrates how unsettled the legal terrain remains.
The case operates in a jurisdiction where copyright doctrine was designed for a world in which reproduction required deliberate copying of specific works — not the ingestion of billions of images into a statistical model. The legal category of training has no clear precedent. Courts are being asked to decide, in effect, whether the pre-digital presumption that unauthorized reproduction is infringement survives into an economy where reproduction has been replaced by pattern extraction.
The lawsuit's significance extends beyond its legal outcome. Whatever the courts decide, Andersen v. Stability AI has functioned as what Thompson would have recognized as a demonstrative action — making visible an institutional deficit that formal governance mechanisms had failed to address. The deficit is specifically that the default arrangement in AI training data acquisition is extraction without consent, and the individual artist has no practical mechanism for preventing or negotiating the extraction.
The case joins the SAG-AFTRA strike and the Authors Guild letter as foundational events in the emerging collective bargaining by code — each improvising from available materials a mechanism of voice for workers excluded from formal AI governance.
Origin
Filed January 13, 2023, in the Northern District of California by Joseph Saveri Law Firm on behalf of Andersen, McKernan, and Ortiz, alongside related suits against Midjourney, DeviantArt, and later Runway.
Key Ideas
Consent grievance. The core claim is that artists' labor was taken without consent to build a tool that devalues that labor.
Legal improvisation. The case uses existing copyright doctrine to address a structural problem the doctrine was not designed for.
Demonstrative function. Beyond its legal outcome, the case makes visible an institutional deficit in the governance of AI training data.
Partial legal progress. Initial dismissal was followed by the 2024 reinstatement of specific claims, leaving key questions unresolved.
Appears in the Orange Pill Cycle
Further reading
- Andersen v. Stability AI Ltd., N.D. Cal. Case No. 3:23-cv-00201 (filed January 13, 2023)
- Matthew Sag, "Copyright Safety for Generative AI" (Houston Law Review, 2023)
- James Grimmelmann, "Copyright for Literate Robots" (Iowa Law Review, 2016)