LAION trains Robert Kneschke photos without consent
Occurred: April 2023
Report incident 🔥 | Improve page 💁 | Access database 🔢
German stock photographer Robert Kneschke discovered that his photos had been used to train the LAION-5B dataset, raising questions about copyright protections from AI datasets and systems, and the ethics of the dataset's eponymous developer.
Having asked LAION to remove his work from their training data, Kneschke received a demand for 887 euros (USD 980) from LAION's law firm Heidrich Rechtsanwälte for what it called an 'unjustified claim' on the basis that it 'only maintains a database containing links to image files that are publicly available on the Internet.'
The fracas put the spotlight on the practices and ethics of LAION, a German-based non-profit dedicated to the 'democratisation' of machine learning research and applications that provides datasets to train major commercial text-to-image and video-generating models such as Stable Diffusion, Midjourney and Google’s Imagen.
Kneschke had used the website Have I Been Trained? to find out whether any major datasets had been trained using his images.
➕ April 2023. Kneschke filed a lawsuit against LAION for copyright infringment.
System 🤖
Operator: LAION
Developer: LAION
Country: Germany
Sector: Media/entertainment/sports/arts
Purpose: Pair text and images
Technology: Database/dataset; Neural network; Deep learning; Machine learning
Issue: Copyright; Ethics
Transparency: Governance; Marketing; Complaints/appeals
Legal, regulatory 👩🏼⚖️
Robert Kneschke v. LAION
Research, advocacy 🧮
Robert Kneschke (2023). LAION-Verein droht Urhebern, die ihre Daten aus KI-Trainingssatz nehmen wollen mit Schadensersatzansprüchen
Campaign for AI Safety (2023). Copyright Cases Against AI Labs
News, commentary, analysis 🗞️
Page info
Type: Incident
Published: August 2023
Last updated: June 2024