Countries Want Royalties for AI

Contents

The Broadcasting Commission of the German states has presented a discussion paper for a "Digital Media State Treaty" (Digitale Medien-Staatsvertrag, DMStV), which provides for comprehensive measures to secure the communicative foundations of a liberal-democratic society. The focus of the initiative, which the broadcasting representatives agreed on on Thursday, is to master the challenges of the AI era.

A central project outlined in the key points is the introduction of an independent, statutory remuneration claim for the use of copyrighted works in the training and deployment of systems for generative AI such as ChatGPT, Gemini, or Claude. The goal is a licensing model that, through collective management by collecting societies, creates fair compensation between innovation and the interests of rights holders, particularly from the journalistic and editorial sector.

At the same time, transparency obligations for AI providers are to be tightened. They would have to disclose in detail which works were specifically used for training large language models as soon as the use goes beyond a mere summary. A clear labeling requirement will also apply to crawlers and bots.

What can be used for AI training?

As part of the most recent major copyright amendment, the EU has defined exceptions to the exclusive exploitation right for text and data mining (TDM). The Bundestag has implemented these provisions in the Copyright Act. Accordingly, the reproduction of legally accessible digital works, for example for training algorithms, is permitted in order to "gain information, in particular about patterns, trends and correlations".

This authorization applies to research institutions, but only under certain conditions: they must not pursue commercial purposes, must reinvest all profits in science, or must act within the scope of a state-recognized mandate in the public interest. This is intended to prevent institutes from extensively mining data in the service of companies.

Considering the increasing spread of generative AI systems, the states want to have it examined whether the existing TDM rules are adequate. An important point is the rights holders' reservation of use: under both national and European law, they are permitted to expressly exclude the use of their works for TDM.

For this blocking mechanism to function effectively and legally, clear formal requirements for the declaration and effect of the reservation are needed, according to the paper. Ultimately, a balanced relationship between copyright protection and the promotion of innovation must be ensured to secure a stable media environment and reliable remuneration structures. Previously, the German Cultural Council had already called for adequate remuneration for the use of protected works by AI. It refers to clear voices stating that the TDM exception does not apply here.

Source references and plausibility check

The states are also pushing for chatbot operators to be held media law responsible, especially if their systems are equivalent to their own content offerings. For this purpose, mandatory source references and links, as well as plausibility checks based on reliable reports, are to be introduced.

In order to strengthen content providers and secure their refinancing, the Broadcasting Commission aims for a balanced regulatory environment. This requires a review of advertising regulations to avoid additional advertising bans, especially on television, and to establish a balance in access to advertising revenue between traditional broadcasting, broadcast-like telemedia, and competing distribution services (platforms).

Providers who work according to journalistic standards and invest in them are to be particularly protected in the online environment. Here, the states are considering concretizing non-discrimination rules, for example, by ensuring that findability is not disadvantaged due to paywalls or external links. At least equal display of independently researched and editorially responsible content compared to exclusively AI-generated content on platforms must be guaranteed.

Do not let Big Tech dictate the rules

To promote the findability of journalistic content, the existing rules for user interfaces are to be evaluated and further developed. The criteria for public added value of reporting are to be sharpened to secure access to reliable content and to provide incentives for corresponding investments. The visibility of media content in search results, feeds, and timelines is to be increased. For this purpose, providers could label reliable content with labels to enable preferential selection by algorithms.

Another focus is on the protection of free communication spaces. This includes protection against manipulative distribution techniques. Practices such as fake accounts, social bots, and non-transparent payment for posts or clicks must be prevented, it states. Furthermore, the safeguarding of editorial independence and transparency is to be improved.

The states are also considering tightening media supervision regarding the deletion of inadmissible content and including further criminal offenses, such as doxxing, rewarding and approving legal violations, or insulting political figures. This is in the Media State Treaty and the Youth Media Protection State Treaty. This is intended to ensure consistency between media regulation and criminal prosecution. The Minister-President of Rhineland-Palatinate, Alexander Schweitzer (SPD), emphasized as chairman of the commission: "We want to develop the rules for our social debates in Europe and Germany ourselves in the digital age, and not let them be dictated by tech giants." First proposals for a DMStV, which primarily concern the implementation of European legal acts, were already made by the states in the summer.

(mma)