Copyright and AI law

AI copyright cases tracker: where every major image lawsuit stands

Two men in a dark interior before shelves of small glowing television screens.
Portrait with television screens.Photograph by Robert Harper. Robert Harper Photograph Library, Archive.London. From the archive, shown for illustration.

Getty, Andersen, Midjourney, LAION, Ross, Anthropic: a dated status table of the cases that will decide how photographs may be used to train AI, separating what courts have ruled from what they have yet to.

Nearly four years after the first lawsuits over image generators were filed, no court in Britain or America has ruled on the question at their centre: whether copying photographs to train an image model infringes copyright. The cases that could answer it are all still running. This page records where each one stands, and is written to be updated as they move.

This article is general information, not legal advice. Checked on 18 November 2026. Each entry separates what a court has decided from what remains pending. Where we could not confirm a point from a court record or a reputable report, we say so.

Image and photography cases: status table

Case Court Decided Pending
Getty Images v Stability AI (UK) High Court; Court of Appeal of England and Wales 4 Nov 2025: model is not an "infringing copy"; limited trade mark infringement. Training claim abandoned, so not ruled on Appeal on secondary infringement heard 11 and 12 Nov 2026. Judgment awaited
Getty Images v Stability AI (US) N.D. California, 3:25-cv-06891 23 Apr 2026: trade mark, dilution and unfair competition claims survive; one claim dismissed with leave to amend Copyright claim over training. Pre-trial; a November 2025 scheduling order set trial for January 2028, according to a docket-based report we have not re-confirmed
Andersen v Stability AI, Midjourney, DeviantArt, Runway N.D. California Aug 2024: copyright and some Lanham Act claims survive dismissal Everything else. Schedule extended Feb and Sep 2026. No trial has taken place
Disney and Universal v Midjourney; Warner Bros. Discovery v Midjourney C.D. California 14 Nov 2025: cases consolidated In discovery. Midjourney agreed on 6 Oct 2026 to share video training datasets. Mediation outcome unverified
Evox Productions v Stability AI and others C.D. California, 2:26-cv-07201 Nothing yet Filed 2 Jul 2026. One defendant moved to dismiss on 12 Oct 2026
Evox Productions v Google N.D. California, 5:26-cv-07013 Nothing yet Filed 9 Jul 2026
Evox Productions v Midjourney C.D. California, 2:26-cv-08408 Nothing yet Filed 30 Jul 2026
Kneschke v LAION Hamburg courts; Federal Court of Justice, I ZR 281/25 27 Sep 2024 and 10 Dec 2025: dataset creation covered by German text and data mining exceptions; natural-language opt-out not machine-readable Heard 3 Sep 2026. No decision reported; a reference to the EU Court of Justice was raised as a possibility
Düsseldorf Higher Regional Court, I-20 W 2/26 Germany 2 Apr 2026, in interim injunction proceedings: an AI image resembling a photograph did not infringe, because protected creative elements were not reproduced Further steps not known

Text and music cases that shape the image disputes

Photographers and agencies watch these because the reasoning will be borrowed.

Case Court Decided Pending
Thomson Reuters v Ross Intelligence D. Delaware; Third Circuit 29 Sep 2026: not fair use to copy Westlaw headnotes to train a non-generative legal search tool. Generative AI distinguished Ross has asked the court to revisit the ruling, according to MLex
Bartz v Anthropic N.D. California 23 Jun 2025: training on lawfully bought books was fair use; keeping pirated books was not. $1.5bn settlement approved 20 Jul 2026 Whether any appeal from approval is pending: unverified
Kadrey v Meta N.D. California 25 Jun 2025: summary judgment for Meta on training, on a limited record about market harm Claims about distribution through BitTorrent continue
New York Times v OpenAI and Microsoft, with consolidated cases S.D. New York No fair-use ruling Summary judgment being briefed; the latest briefing dates reported run to 18 Dec 2026
Concord v Anthropic N.D. California No ruling Fair-use summary judgment argued 21 Oct 2026 and taken under submission
GEMA v OpenAI Munich Regional Court I, 42 O 14139/24 11 Nov 2025: memorised song lyrics are reproductions outside the text and data mining exception On appeal to the Munich Higher Regional Court
GEMA v Suno Munich Regional Court I, 42 O 763/25 31 Jul 2026: training on stream-ripped music infringed; exception failed partly for lack of lawful access Not final. Appeal status unverified
Like Company v Google Ireland EU Court of Justice, C-250/25 Nothing binding. Advocate General's opinion delivered 3 Nov 2026 Judgment, date not set

Timeline of key dates

Date Event
January 2023 Getty issues its High Court claim against Stability AI; Andersen class action filed in California
27 December 2023 New York Times sues OpenAI and Microsoft
August 2024 Andersen: core claims survive motion to dismiss
27 September 2024 Kneschke v LAION: Hamburg Regional Court rules for LAION
11 February 2025 Thomson Reuters v Ross: district court rejects fair use
11 June 2025 Disney and Universal sue Midjourney
9 to 30 June 2025 Getty v Stability AI trial in London; training claim abandoned
23 and 25 June 2025 Bartz v Anthropic and Kadrey v Meta fair-use rulings
14 August 2025 Getty refiles its US claim in the Northern District of California
4 November 2025 Getty v Stability AI: High Court judgment
11 November 2025 GEMA v OpenAI: Munich judgment
10 December 2025 Kneschke v LAION: Hamburg appeal dismissed
2 April 2026 Düsseldorf appeal ruling on an AI image resembling a photograph
23 April 2026 Getty v Stability AI (US): motion to dismiss decided
July 2026 Evox Productions files three image suits
20 July 2026 Bartz settlement receives final approval
31 July 2026 GEMA v Suno: Munich judgment
3 September 2026 Kneschke v LAION: Federal Court of Justice hearing
29 September 2026 Thomson Reuters v Ross: Third Circuit affirms
21 October 2026 Concord v Anthropic: summary judgment hearing
3 November 2026 Like Company v Google: Advocate General's opinion
11 and 12 November 2026 Getty v Stability AI: Court of Appeal hearing

What has actually been decided?

Less than the volume of litigation suggests. Drawing only on the rulings above:

  • In England, a trained model that does not store the training images is not an "infringing copy" for the purposes of secondary infringement. That finding is the one under appeal. The detail is in our Getty v Stability AI timeline.
  • In Germany, building a research dataset of image links and descriptions was held to fall within the text and data mining exceptions, and a reservation written in ordinary language did not count as machine-readable. That too is under appeal; see Kneschke v LAION and machine-readable opt-outs.
  • Also in Germany, a model that has memorised a work and can reproduce it is outside the exception, and so is training on unlawfully obtained copies. Both rulings concern music and neither is final.
  • In the United States, three courts have ruled on fair use for training on text. Two found for the developer on particular facts and one, on appeal, against. Our article on fair use after Thomson Reuters v Ross compares them.

What is still pending?

The decisions most likely to change practice for photographs, in no predicted order:

  1. The Court of Appeal's judgment in Getty v Stability AI.
  2. The Federal Court of Justice's decision in Kneschke, or a reference from it to the EU Court of Justice.
  3. The EU Court of Justice's judgment in Like Company v Google Ireland.
  4. A first US fair-use ruling on an image model, in Getty, Andersen, the Midjourney cases or the Evox suits.
  5. Summary judgment in the New York Times and Concord cases.

This page covers the cases with the most bearing on images and training data. Many more copyright suits against AI companies are on foot, particularly in the United States.

Claims to treat with caution

Some statements circulate widely and are not supported by the record as we found it.

  • "Getty decided that AI training is lawful in the UK." The training claim was abandoned. The court did not rule on it.
  • "The Andersen case went to a jury in September 2026." Docket-based reports show the schedule was extended that month. No trial has been held.
  • "AI training is fair use in the US", and its opposite. Neither is established. The decided cases concern text, turn on their facts, and no court has ruled on image models.
  • "The Bartz settlement sets a price per work." It resolved claims over pirated books under statutory-damages exposure. It is not a licence rate and does not concern photographs.
  • "A notice in website terms is a valid EU opt-out." The Hamburg appeal court held otherwise, and the point is before the Federal Court of Justice.

How to follow these cases yourself

Primary records are the safest source. English judgments are published by the National Archives' Find Case Law service. EU opinions and judgments appear on the Court of Justice's Curia site. US dockets can be read through CourtListener. Law-firm trackers such as Mishcon de Reya's are useful for cross-checking, and the better ones date every entry.

If your interest is practical, the working documents matter as much as the judgments. Knowing exactly which files were supplied to whom is the foundation of any later claim or defence, and a fingerprint of each file is the simplest record. Our free image hash and fingerprint tool generates one in the browser.

The alternative to waiting

Every case in these tables arose from a dispute over use said to have taken place without a licence. For buyers of image data, the more certain route is the contractual one: a licence from the owner, on stated terms, for a described set of images.

Archive.London is the home of the Robert Harper Photograph Library, 95,487 unique photographs on the inventory's count by a single photographer, owned by 825 Ventures Ltd, the owner and rights holder. More than half of the archive is still on film and has yet to be scanned, and only a small selection has been published online by Archive.London, from studio still life to the music collections. Licences are negotiated directly and are subject to rights review. Counsel, journalists and data teams who want to understand what is available can start at AI training data licensing.

Frequently asked questions

Not directly, as of 18 November 2026. The English High Court did not rule on training in Getty v Stability AI because that claim was abandoned. German courts held that building the LAION dataset was covered by text and data mining exceptions. No US court has decided fair use for an image model. Decisions on text and music exist and point in different directions.

What is the status of Andersen v Stability AI?

It is still in the pre-trial phase in the Northern District of California. Copyright claims and some Lanham Act claims survived a motion to dismiss in August 2024. The schedule was extended in February 2026 and again in September 2026. No trial has taken place, despite at least one widely indexed post claiming otherwise.

What is happening in the Disney and Universal lawsuit against Midjourney?

The case, filed on 11 June 2025 and consolidated with Warner Bros. Discovery's suit on 14 November 2025, is in discovery in the Central District of California. Midjourney agreed on 6 October 2026 to hand over video training datasets. A court-ordered mediation deadline passed on 19 August 2026; we could not verify the outcome.

Getty Images v Stability AI in London and California; Kneschke v LAION in Germany, brought by a stock photographer; the three Evox Productions suits filed in July 2026 by an automotive photo library; and Andersen v Stability AI, which concerns visual artists more broadly. A German appeal ruling of 2 April 2026 also considered an AI image resembling a photograph.

How often is this tracker updated?

The page carries a 'checked on' date and each row says whether the point is decided or pending. This version was checked on 18 November 2026. Litigation moves quickly and court schedules change, so confirm the current position against the court record or a recent report before relying on any entry.

Sources

This article is general information, not legal advice. It reflects the position as we understood it on the date shown and may be out of date. Published by Archive.London. Corrections are welcome under our editorial policy.

Licensing the archive for AI and data projects

95,487 photographs from one photographer and one rights holder. Read the dataset card, search the 503-folder inventory, discuss a sample set under NDA, or arrange a first look as scanning proceeds.

See AI and data licensingWrite to the archive