Getty's US filing against Stability AI was a separate suit

The image

The dispute concerns not one photograph but a claimed pool of them: Getty Images' amended complaint against Stability AI states that, 'upon information and belief,' Stability AI copied more than 12 million photographs from Getty's collection, along with their captions and metadata, to build a competing image-generation system. The complaint is Getty's own pleading, not a court's finding, and it identifies the underlying images only in the aggregate, by reference to Getty's licensed collection and to exhibits listing registered works, rather than by describing any single frame.

What the documents show

Court records show Getty Images (US), Inc. filed suit against Stability AI, Inc. in the U.S. District Court for the District of Delaware on 3 February 2023, docketed as No. 1:23-cv-00135, and filed an amended complaint on 29 March 2023 adding Stability AI, Ltd. as a defendant. The amended complaint pleads copyright infringement, false and removed copyright-management information under 17 U.S.C. § 1202, and trademark infringement and dilution under the Lanham Act and Delaware law. This filing is separate from the UK proceeding: Getty's own January 2023 statement describes a distinct claim filed days earlier 'in the High Court of Justice in London.' The Delaware docket shows Getty filed a notice of voluntary dismissal on 14 August 2025, closing the case on 18 August 2025; no document opened for this entry records a merits ruling as of the retrieval date.

Production context

For an image library researching AI-image provenance, the complaint's own account illustrates what a licensing dispute over training data looks like on paper before any court tests it: an assertion of scraped, uncompensated use of a commercial archive's assets, tied to metadata and copyright-management fields the plaintiff says were altered. This is a useful record of how a rights holder frames a training-data claim; it is not evidence the allegations were proven or accepted by any court, since the case closed by voluntary dismissal rather than adjudication.

Reference versus imitation

Reference practice here means citing the complaint precisely as an allegation, attributed to Getty, filed on a specific docket, rather than restating its claims as established fact. It would be imitation of the litigation record itself to describe Stability AI as having infringed Getty's images without qualification, when the case record available shows only a filed and later voluntarily dismissed complaint, with no cited ruling on the merits. The distinction between a party's pleading and a court's finding is the whole of what the record here can support.

• Is a given claim about this case sourced to the complaint, to a ruling, or to neither?
• Does a summary distinguish the Delaware federal case from Getty's separate UK proceeding?
• What, if anything, does the docket show happened after the case was voluntarily dismissed?

The Delaware case remains, on this record, an allegation and a docket history rather than a decided dispute. Any use of it as a reference point for AI training-data practice should say so, and should keep Getty's pleaded figures and legal theories clearly attributed to Getty's own complaint.