A European View on Generative AI and Copyright

• Pages: 23-29
• Date: 01 June 2025
• Published date: 01 June 2025
• Author: Marcus Von Welser

LANDSLIDE June/July 2025

Published in Landslide, Volume 17, Number 4, 2025. © 2025 by the American Bar Association. Reproduced with permission. All rights reserved. This information or any portion

thereof may not be copied or disseminated in any form or by any means or stored in an electronic database or retrieval system without the express written consent of the

American Bar Association.

23

A European View on

Generative AI and Copyright

Marcus von Welser

T

he development of artif‌icial intelligence (AI) has accel-

erated in recent years, mainly due to the amount of

data available and the use of multilayer artif‌icial neural

networks.1 One application of artif‌icial neural networks is natu-

ral language processing.

The technology used by ChatGPT is based on a generative

pretrained transformer (GPT) language model. These are based

on the transformer architecture introduced by Google in 2017.2

Transformers are specialized artif‌icial neural networks that

allow large amounts of text to be processed in parallel.3 Artif‌icial

neural networks consist of input, output, and hidden layers of

artif‌icial neurons. There are weighted connections called param-

eters between the neurons. For example, GPT-3 had 175 billion

weighted connections between the artif‌icial neurons.

4

The numeri

-

cal values of these parameters are determined by training and

continuously adjusted. Training is nothing more than adjusting the

weights. Since large language models are trained with texts, some

of which are copyrighted, and can also reproduce them, there is

a tension with copyright law. The questions that arise are of both

a legal and an ethical nature.

On December 27, 2023, the New York Times Co. f‌iled a lawsuit

against OpenAI and Microsoft for copyright infringement.5 In the

lawsuit, the New York Times accuses the defendants of using millions

of New York Times articles to train large language models without

authorization. The New York Times alleges in its complaint that it

owns the copyrights to more than three million articles and seeks,

among other things, an order that the defendants pay damages,

cease and desist from further copyright infringement, and destroy

Marcus von Welser is a partner in the German intellectual

property law firm Vossius & Partner. He advises and

represents clients in all areas of intellectual property law, in

particular copyright, trademark, and unfair competition law.

He has extensive experience in both litigation and contract

negotiation. He can be reached at m.welser@vossius.eu.

Image: Getty Images/Bloomberg Creative