State of the proceedings
The LAION case: the machine-readable reservation from Hamburg to the Federal Court of Justice
As of 25 September 2026. This page is updated when the proceedings move.
A photographer sued a non-profit association, LAION, that had downloaded a preview of one of his photographs while building a dataset for training AI models. The case is now before the Federal Court of Justice, and it turns in part on whether a reservation was machine-readable at the time the image was copied. This page records the course of the case and what the published sources say, with each source named.
Course of the case
- Regional Court of Hamburg, judgment of 27 September 2024, case no. 310 O 227/23: action dismissed.
- Higher Regional Court of Hamburg, judgment of 10 December 2025, case no. 5 U 104/24: appeal dismissed.
- Federal Court of Justice, First Civil Senate, case no. I ZR 281/25: oral hearing on 3 September 2026. The Court has scheduled the pronouncement of its decision for 17 December 2026, 8:45.
Source: Federal Court of Justice, press release no. 085/2026 of 20 May 2026, and the Court's notice of the pronouncement date for I ZR 281/25 (bundesgerichtshof.de).
The facts, as the Court describes them
According to the press release, the association makes a dataset of 5.85 billion image-text pairs publicly available free of charge. The dataset holds links to images and a description of each image, not the images themselves. To build it, the association downloaded the images, checked with software whether each description matched its image, and kept the metadata of the images that did. The downloads were made in the second half of 2021. One of them was the preview image of the photographer's work on the website of a picture agency, whose website prohibited access by “automated programs, applets, bots or the like”.
What the court of appeal held on the reservation
The Higher Regional Court held that the download was a reproduction, but that it was covered by the exceptions in Sections 44b and 60d of the German Copyright Act. On the reservation, the Federal Court of Justice summarises the court of appeal's reasoning as follows:
“A reservation of use is, under Section 44b(3) UrhG, effective only if it is made in machine-readable form. The claimant had not shown that the reservation, drafted in natural language, was machine-readable at the relevant time of use in the second half of 2021.”
Federal Court of Justice, press release no. 085/2026, summarising the judgment of the Higher Regional Court of Hamburg of 10 December 2025. Our translation; only the German text is authoritative. Source: bundesgerichtshof.de.
As a second, independent ground, the court of appeal held the use to be covered by Section 60d, the exception for text and data mining for the purposes of scientific research, and regarded the association as a research organisation within the meaning of that provision.
The wording of Article 4(3) of the DSM Directive and of Section 44b is set out on the page on text and data mining and the machine-readable reservation.
What is not yet known
The Federal Court of Justice has not yet decided. Until the decision is pronounced, what the Court will hold, and whether it will refer questions to the Court of Justice of the European Union, is not known. This page reports decisions and dates once they are published by the courts themselves.
The point in time
The passage above names a point in time: the second half of 2021, when the image was copied. Where a reservation is at issue, the question of fact is what a website presented to machines on a particular day. Servers keep no history of that. robots.txt and similar files are overwritten when they change, and a copy taken later shows the later day.
This archive records those files every day for around 128,000 EU domains, since 22 July 2026. It holds nothing from before that date. What we store lists every file; the coverage check shows whether a domain is observed and since when.
Not legal advice. This page reports the course of court proceedings and quotes the courts' own publications with their sources. It does not assess the decisions, predict their outcome or say what anyone should do. MachineWitness is operated by a software company, not a law firm; for the assessment of a specific situation, ask a lawyer.