“Fundamentals are threatened!” With this announcement, an alliance of authors and artists from 43 European associations under the aegis of the copyright initiative is warning of recent developments in the field of artificial intelligence (AI). “AI-generated products have a direct impact on social life,” says a statement to European and German policymakers published on Wednesday. The technology may only come onto the market under strict conditions.
To the signatories of the appeal include the journalists’ and authors’ associations DJV and PEN, the photographers’ association Freelens, the Federal Acting Association and trade unions such as DGB and Verdi. Together with companies from the book, music and film industries as well as from radio and the press, they refer to the “impact of generative AI in particular”.
“Disinformation and Manipulation”
By this, the signatories mean AI software such as ChatGPT, Stable Diffusion or DALL-E, which generate texts, images, music and videos based on existing, often copyrighted works. With these applications, the alliance makes an “immanent potential for disinformation and manipulation” and underlines the concern expressed “even by AI experts” about a loss of control and their “calls for legal limits”.
The signatories warned of “damage to European society, the economy and culture” and expressed concern that “some in politics” saw no need for action. They welcome the fact that the EU Parliament wants to set stricter rules for “General Purpose AI” with the planned AI regulation, but consider these to be insufficient “to adequately protect the digital ecosystem and society as a whole”.
“Regulate the entire chain of origin”
The call is therefore: “Generative AI must be regulated along its entire creation chain.” Protected works are used “to train AI, often without permission and without compensation, and not always for legitimate purposes”. In the future, models used will only be allowed to be brought to European markets if the training material is transparent and sufficiently resilient “in terms of correctness, accuracy, objectivity and variety”.
According to the demand, model providers such as Google or Meta should be prohibited from operating central platform services for the distribution of digital content within the meaning of the Digital Markets Act (DMA). Part of the ongoing operation of an AI system must also “take place via a computing infrastructure stationed in Europe”. In a second step, the EU legislators should make it clear that the EU exceptions for text and data mining are not linked to permission to “replace sources without any remuneration”.
(vbr)
