Hearing scheduled for September 3, 2026 at 9:00 a.m. in case I ZR 281/25 (creation of a data set for AI training)
Year of issue 2026
Release date May 20, 2026
No. 085/2026
The First Civil Senate of the Federal Court of Justice, which is responsible, among other things, for copyright law, has to decide whether the reproduction of a photograph in the creation of a data set for the training of generative artificial intelligence (AI) infringes the copyright of the photographer.
Facts:
The plaintiff is a photographer. The defendant is a registered non-profit association whose stated purpose is to promote research and education. The defendant makes a dataset of image-text pairs publicly available free of charge. This dataset contains hyperlinks to images publicly accessible on the internet, as well as further information about the corresponding images, including a textual description of the image's content. The defendant's dataset comprises 5.85 billion such image-text pairs and can be used to train so-called generative artificial intelligence.
To create this dataset, the defendant used an existing dataset from the USA, which contained the web addresses (URLs) of images found on the internet and a description of the respective image content. The defendant extracted the URLs to the images from this dataset and downloaded the images from their respective storage locations. Subsequently, the defendant used software to check whether the description of the image content already present in the existing dataset actually matched the content of the image. Images where the text and image content did not sufficiently match were filtered out. For the remaining images, the metadata, in particular the URL of the image's storage location and the image description, were extracted and included in a newly created dataset of the defendant. The images themselves are not included in the defendant's dataset. The downloads necessary for creating the defendant's dataset were carried out in the second half of 2021. This also included a photograph taken by the plaintiff and the download of the preview image of this photograph, which was posted on the website of a stock photo agency. The photo agency had prohibited access to its website by "automated programs, applets, bots or the like" on its website.
The plaintiff argues that the defendant's reproduction of his photograph during the analysis process infringes his copyright. He is suing the defendant for an injunction.
Previous course of proceedings:
The regional court dismissed the lawsuit. The plaintiff's appeal was unsuccessful.
The Court of Appeal reasoned that, by downloading the preview image, the defendant had indeed reproduced the plaintiff's photograph in accordance with Section 16 Paragraph 1 of the German Copyright Act (UrhG), without obtaining the plaintiff's consent or a license from his image agency. However, this use of the image was justified by the limitations on copyright stipulated in Sections 44b and 60d of the German Copyright Act (UrhG).
According to Section 44b of the German Copyright Act (UrhG), reproductions of lawfully accessible works for the automated analysis of one or more digital or digitized works are permitted in order to obtain information, in particular about patterns, trends, and correlations (so-called text and data mining), provided the copyright holder has not reserved these uses. The plaintiff's photograph, which is available online, is a lawfully accessible digital work, the reproduction of which served the purpose of automated analysis, namely, comparing the image content with the accompanying image description using software. Determining whether a specific image and its description match constitutes information within the meaning of Section 44b Paragraph 1 of the German Copyright Act. Since the comparison between the image and its description already constitutes an analysis for obtaining information within the meaning of this provision, it is irrelevant whether the subsequent training of generative AI models also fulfills the requirements of text and data mining. The image agency's reservation of rights does not preclude the use of the photograph created by the plaintiff. A reservation of rights of use is only effective under Section 44b Paragraph 3 of the German Copyright Act if it is in machine-readable form. The plaintiff has not demonstrated that the reservation, written in natural language, was machine-readable at the relevant time of use in the second half of 2021.
The use of the photograph is also covered by the limitations on copyright under Section 60d Paragraph 1 of the German Copyright Act (UrhG), according to which reproductions for text and data mining for scientific research purposes are permissible under certain conditions. The defendant is a research organization within the meaning of Section 60d Paragraph 2 Sentence 2 No. 1 of the German Copyright Act (UrhG). The creation of the dataset constitutes a methodical, verifiable process aimed at generating future knowledge, which is attributable to applied research. The fact that the dataset can also be used by providers of commercial AI models does not contradict this. It is inherent in the open-source approach pursued by the defendant that the results of its work are available to everyone, including commercial providers.
According to the three-step test to be carried out pursuant to Article 5(5) of Directive 2001/29/EC on the harmonisation of certain aspects of copyright and related rights in the information society, there is also no undue infringement of the plaintiff's interests as the rights holder.
With his appeal, which was granted by the Court of Appeal, the plaintiff continues to pursue his claim.
Lower courts:
Hamburg Regional Court - Judgment of September 27, 2024 - 310 O 227/23
Hamburg Higher Regional Court - Judgment of December 10, 2025 - 5 U 104/24
The relevant regulations are as follows:
Section 16 Paragraph 1 Copyright Act
The right of reproduction is the right to produce copies of the work, whether temporary or permanent, by any means and in any number.
Section 44b Copyright Act (excerpt)
(1) Text and data mining is the automated analysis of one or more digital or digitized works to obtain information, in particular about patterns, trends and correlations.
(2) Reproductions of lawfully accessible works for text and data mining are permitted. (...)
(3) Uses pursuant to paragraph 2 sentence 1 are only permitted if the copyright holder has not reserved such rights. A reservation of rights of use for works accessible online is only effective if it is made in machine-readable form.
Section 60d Copyright Act (excerpt)
(1) Reproductions for text and data mining (Section 44b paragraphs 1 and 2 sentence 1) are permitted for scientific research purposes in accordance with the following provisions.
(2) Research organisations are entitled to make copies. Research organisations are universities, research institutes or other institutions that conduct scientific research, provided that they
1. pursue non-commercial purposes,
2. (...)
Article 5(5) of Directive 2001/29/EC (excerpt)
The (...) exceptions and limitations may only be applied in certain special cases where the normal exploitation of the work or other protected subject matter is not impaired and the legitimate interests of the rights holder are not unduly violated.
Karlsruhe, May 20, 2026
Press Office of the Federal Court of Justice
, 76125 Karlsruhe,
Telephone (0721) 159-5013,
Fax (0721) 159-5501