Public Beta · Not legal advice
← Interactive map

Germany · DE-001

Kneschke v LAION

HELDLG Hamburg; Hanseatisches OLG; Bundesgerichtshof310 O 227/23; 5 U 104/24; I ZR 281/25Revision pending before BGH
Text and Data MiningRights Reservation / Opt-OutReproductionTraining Data Acquisition

Key holding or allegation

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.

Why it matters

Leading European national authority on dataset-preparation TDM and opt-out formalities in the generative-AI ecosystem.

Facts

LAION downloaded a photograph while analysing image-text correspondence in creating an image-text dataset that could later be used for generative-AI training.

Issues before the court

Whether the reproduction made during dataset preparation was covered by German TDM exceptions and whether a rights reservation was machine-readable.

What the court did not decide

The courts did not establish a categorical rule that subsequent foundation-model training is lawful TDM; the BGH description itself distinguishes dataset creation from later training.

Procedural timeline

2024-09-27

LG Hamburg
First-instance merits judgment: Claim dismissed under the German TDM/research framework for copying used in automated image-text analysis during dataset creation.

HELD
2025-12-10

Hanseatisches OLG
Appellate merits judgment: Appeal dismissed; section 44b TDM held applicable to the image-analysis activity and natural-language reservation held insufficiently machine-readable; section 60d also supported the result.

HELD
2026-09-03 scheduled

Bundesgerichtshof
Hearing scheduled: BGH hearing scheduled on the copyright/TDM issues arising from copying a photograph during creation of an AI-training-related dataset.

QUESTION REFERRED