Case inventory · Public Beta
The inventory separates matters from their decisions and procedural events. Foundational and analogical authorities are excluded. This Beta resource may contain omissions or errors and must not be relied upon as legal advice. Read the disclaimer.
Beijing Internet Court
The court did not attribute authorship to the software-generated text as such under the natural-person authorship framework, while recognising protectable human-created aspects of the overall report.
View matter →Shenzhen Nanshan District People's Court
The court recognised protection based on the human-organised process of selection, analysis, judgment and arrangement underlying the generated article.
View matter →Beijing Internet Court
The court held that the user's iterative prompting, parameter choices, adjustment and selection were sufficient on the facts for copyright protection and human authorship.
View matter →Shanghai Jiading District People's Court
The court found that face replacement did not transform the underlying protected audiovisual expression into a sufficiently independent new work and found infringement in the commercial dissemination context.
View matter →Guangzhou Internet Court
The court found infringing substantially similar generated images and imposed cessation, technical preventive measures and damages.
View matter →Hangzhou Internet Court; Hangzhou Intermediate People's Court
The courts separated user conduct from platform conduct and upheld liability based on the platform's knowledge, commercial model and failure to take reasonable measures, while addressing cessation of infringing model availability.
View matter →Changshu People's Court
The court recognised copyright in the final image based on the demonstrated human creative process and editing.
View matter →Wuhan East Lake High-Tech Development Zone People's Court
The court found sufficient human creative choices on the facts to recognise copyright protection.
View matter →Zhangjiagang People's Court; Suzhou appellate court
Protection failed on the evidence because the claimant did not sufficiently establish the generation process and individual human creative contribution.
View matter →Shanghai Huangpu District People's Court
The court treated the list-style prompts predominantly as functional instructions or ideas and found insufficient original literary expression on the facts.
View matter →Shanghai IP Court
Reported appellate reasoning distinguished LoRA statistical parameterisation from reproduction of the training image itself while affirming infringement where the trained model was used to generate substantially reproducing outputs; platform liability was rejected on the reported facts.
View matter →Beijing Internet Court
Plaintiffs allege unauthorised training use; no merits holding has been verified for this snapshot.
View matter →U.S. District Court for D.C.; U.S. Court of Appeals for the D.C. Circuit; U.S. Supreme Court
The D.C. Circuit held that copyrightable works must be authored in the first instance by a human being.
View matter →U.S. District Court for the District of Colorado
No merits judgment had been identified by the project cutoff; plaintiff argues that prompting and post-processing satisfy human authorship.
View matter →U.S. District Court for the Northern District of California
The court allowed significant training/model infringement theories to proceed at the pleading stage, accepting allegations as plausible without finding that model weights actually contain infringing copies or that training infringes.
View matter →U.S. District Court for the Northern District of California; U.S. Court of Appeals for the Ninth Circuit
The matter produced substantive CMI rulings and an interlocutory Ninth Circuit appeal; final appellate resolution had not been identified by the cutoff.
View matter →U.S. District Court for the Southern District of New York
An April 2025 ruling allowed important contributory-infringement theories to proceed and treated different CMI theories separately.
View matter →U.S. District Court for the Southern District of New York
The section 1202(b)(1) removal claim against OpenAI survived while other CMI theories and Microsoft were dismissed.
View matter →U.S. District Court for the Southern District of New York
The action was dismissed for lack of Article III standing, and reconsideration was denied.
View matter →U.S. District Court for the Northern District of California
The court held the LLM training uses before it highly transformative and fair use, and treated digitisation of purchased books as fair use, while analysing pirated acquisition/retention separately and refusing fair use for that conduct on the summary-judgment record.
View matter →U.S. District Court for the Northern District of California
Meta obtained summary judgment on the named plaintiffs' training claim because the evidentiary record did not establish sufficient market harm, despite the court recognising that generative-AI market dilution can be legally important.
View matter →U.S. District Court for the Northern District of California
The preliminary injunction was denied principally for failure to establish irreparable harm; the court expressly left the training infringement/fair-use merits unresolved.
View matter →U.S. District Court for the Southern District of New York
An October 2025 decision held that consolidated plaintiffs adequately pleaded an output-based copyright infringement claim; this was not a finding of actual infringement.
View matter →U.S. District Court for the Southern District of New York
Complaint allegations only; no merits ruling.
View matter →U.S. District Court for the Southern District of New York
Direct-output and secondary-liability theories survived the pleading stage.
View matter →U.S. District Court for the Southern District of New York
Perplexity's motion to dismiss or transfer on threshold grounds was rejected; copyright merits remain allegations.
View matter →U.S. District Court for the Southern District of New York
Some copyright theories concerning use of the original recordings survived, while broader theories based simply on training the voice generator were dismissed at the pleading stage.
View matter →U.S. District Court for the Northern District of California
A September 2025 ruling allowed claims tied sufficiently to specified models to continue while dismissing claims concerning models for which the complaint did not plausibly connect plaintiffs' works to training.
View matter →U.S. District Court for the Northern District of California
The court found allegations insufficient where plaintiffs could not plausibly connect their specific works to the actual training data of the challenged models.
View matter →U.S. District Court for the Northern District of California
Significant training claims survived where plaintiffs sufficiently connected works to alleged training materials; a contributory theory based on downloading/processing scripts also survived, while vicarious liability was narrowed.
View matter →U.S. District Court for the Southern District of New York
In April 2026 the court declined to dismiss the DMCA section 1201 circumvention claim concerning alleged acquisition of training material.
View matter →U.S. District Court for the Southern District of New York
A May 2026 ruling allowed parts of the artists' case to continue while dismissing other theories.
View matter →U.S. District Court for the Northern District of California
The court held the alleged BitTorrent direct/vicarious/contributory infringement theories sufficient to proceed; proof that a film reached a model was not necessary to plead the alleged torrent infringement itself.
View matter →U.S. District Court for the Northern District of California
The April 2026 order dismissed the section 1202(a) false-CMI theory without prejudice while allowing other related claims to proceed; the underlying training infringement claim was not resolved by that order.
View matter →U.S. District Court for the Central District of California
Complaint allegations only for the core copyright merits; no substantive merits judgment identified by the cutoff.
View matter →U.S. District Court for the Central District of California
On 22 May 2026 the court denied motions to dismiss, finding the studios' training and near-identical output allegations sufficiently plausible to proceed.
View matter →U.S. District Court for the Southern District of New York
Complaint allegations only; no merits ruling identified by the cutoff.
View matter →U.S. District Court
Complaint allegations only; no merits ruling by the cutoff.
View matter →U.S. District Court for the District of Massachusetts
Core merits remained unresolved in the project snapshot; party-specific business settlements/licensing developments require careful separation from the court record.
View matter →LG Hamburg; Hanseatisches OLG; Bundesgerichtshof
LG Hamburg dismissed the claim relying on the research TDM framework; OLG Hamburg affirmed and held section 44b TDM applicable to the image-analysis activity, finding the natural-language reservation insufficiently machine-readable, with section 60d also supporting the result.
View matter →LG Munchen I; appeal to OLG Munchen
LG Munchen I held that reproducibly memorised protected lyrics in GPT-4/4o constituted copies embodied in the model, that TDM did not justify the persistent model-level reproduction found, and that specified outputs separately infringed.
View matter →LG Munchen I
The court found protected compositions reproducibly contained in Suno models and found infringing outputs; it also applied U.S. law to U.S. training acts and rejected fair use on the facts presented.
View matter →AG Munchen
The court found no protection on the evidence because the prompts left the actual creative decisions to the AI, while recognising that copyright may be possible where human creative choices objectively determine the output.
View matter →LG Dusseldorf; OLG Dusseldorf
The appellate court found no sufficient basis for treating the AI output as a new human-authored work but also found no copyright infringement because only unprotected motif elements, not the photographer's protected creative expression, were reproduced.
View matter →LG Frankfurt am Main
The court treated the human-written lyrics as protectable despite AI use in producing the song and focused on evidence distinguishing human-created from AI-generated elements.
View matter →LG Munchen I
Plaintiff allegations only; no merits decision identified by the cutoff.
View matter →High Court of Justice, Chancery Division; Court of Appeal
The High Court held that an intangible AI model can in principle be an article, but Stable Diffusion was not an infringing copy on the surviving copyright theory where the model did not itself contain or store copies of Getty's protected works as the case was presented.
View matter →Prague Municipal Court
Claims were dismissed because the claimant failed to prove the asserted prompting/generation process and because the court considered the AI-generated image not to be the unique creative result of a natural person.
View matter →High Court of Delhi at New Delhi
At the interim stage the Delhi High Court held prima facie that storing publicly accessible ANI works for LLM training fell within section 52(1)(a) fair dealing for private use including research, and ANI had not established substantial reproduction or memorisation in the tested outputs.
View matter →Madras High Court
The Madras High Court granted short-term interim restraints including the requested AI-related forms of use on a prima facie record.
View matter →High Court of Delhi
The court declined immediate ex parte protection because it was not yet satisfied as to copyright subsistence in the overall work given the AI-generated musical composition.
View matter →Tribunal de Justica de Sao Paulo
The appellate court overturned the dismissal and required further evidentiary proceedings, recognising that use of generative AI does not by itself eliminate possible infringement of related or personality rights.
View matter →Tribunal de Justica de Santa Catarina
The appellate court refused to suspend royalties, stressing the factual/legal complexity and possible reliance on protected works; lack of an identifiable human author did not itself establish that charges were inapplicable.
View matter →Ontario Superior Court of Justice, Commercial List
The Ontario court retained jurisdiction over six OpenAI entities while dismissing four insufficiently connected entities; the court expressly did not determine the merits of the publishers' copyright allegations.
View matter →Tokyo District Court
Plaintiff allegations only; first oral hearing reported 26 May 2026 and no merits judgment identified by the cutoff.
View matter →Tokyo District Court
Plaintiff allegations only; Perplexity contested the claims at the first oral hearing and no merits decision was identified by the cutoff.
View matter →Tribunal judiciaire de Paris, 3rd Chamber
Plaintiff allegations only; SGDL confirmed in 2026 that the infringement proceeding remained pending before the Paris Judicial Court.
View matter →Civil Court of Rome
Plaintiff allegations only; no merits judgment identified by the cutoff.
View matter →Danish court - exact public docket not secured
Koda's allegations only; no Danish merits ruling identified.
View matter →Federal Administrative Justice Tribunal; Supreme Court of Justice of the Nation
The Supreme Court upheld the refusal of protection/registration under the natural-person authorship framework on the facts presented.
View matter →Copyright Tribunal of Kenya
The Tribunal discussed the need for sufficient human creative intervention but did not finally determine copyrightability of the particular works because the evidentiary record was insufficient; it set aside the Copyright Board decision on jurisdictional grounds.
View matter →Budapest Kornyeki Torvenyszek
No domestic merits holding; the Hungarian court referred EU-law questions to the CJEU.
View matter →Court of Justice of the European Union
Questions referred only. The CJEU had not delivered a merits judgment by 18 Aug 2026.
View matter →