A research paper commissioned by the European Parliament has called for an EU law to pay writers, musicians, and artists whose work has been used to train GenAI models.
Controversy surrounds large language models and other GenAI systems because they rely on vast swathes of copyrighted material as training data without adequate remuneration of the people and organizations creating the content.
The study finds [PDF] the current exception for text-and-data mining (TDM) in EU law “was not designed to accommodate the expressive and synthetic nature of generative AI training, and its application to such systems risks distorting the purpose and limits of EU copyright exceptions.”
Companies backing GenAI have argued that the use of copyrighted training data is defensible under “fair use,” similar to a student using a copyrighted book to acquire knowledge.
But the new research paper – commissioned by the European Parliament’s Policy Department for Justice, Civil Liberties and Institutional Affairs at the request of the Committee on Legal Affairs – found that the argument doesn’t stand up.
“While it is often suggested that AI systems ‘learn’ in ways similar to humans – such as reading a book or studying a painting – this analogy is misleading from a legal perspective,” the paper said. “Under EU copyright law, this study finds that such a comparison does not hold. When generative AI models are trained on protected content, they typically make copies and process the actual expressions found in those works. This goes beyond what is permitted under current legal exceptions for activities like research or analysis.”
It argued that AI systems do not “understand” what they process in the same way humans do. “As philosopher Luciano Floridi puts it, AI acts without understanding – it follows statistical patterns rather than engaging with meaning. This difference matters legally,” it said.
The paper proposes a new EU-level statutory exception to copyright for the specific purpose of training generative AI systems, but also the introduction of an “unwaivable right to equitable remuneration for authors and rightsholders whose works are used in such training.”
It also argues that while fully machine-generated outputs should remain unprotected, AI-assisted works require “harmonised protection criteria.”
In May, the head of the US Copyright Office, Shira Perlmutter, was removed from office after her agency concluded that AI developers’ use of copyrighted material exceeded the bounds of existing fair use doctrine. Her report argued that fair use does not cover the commercial use of large volumes of copyrighted works to generate expressive content that competes in existing markets.
She is currently appealing against her dismissal.
The current status quo in the US is being challenged by a number of lawsuits. Mega studios Disney and Universal are suing GenAI provider Midjourney for what they claim is a “bottomless pit of plagiarism.”
The claim alleges that Midjourney copied some of the studios’ best-known characters from their vast movie portfolios. ®