AI should pay! Countries demand copyright levy for AI systems

No more free of charge: Those who work with content should also pay for it

Human-generated content such as text, images or music - all of this now flows into huge AI models such as ChatGPT, Gemini or Claude. And on a massive scale. The systems learn, analyze and produce - in many cases with copyrighted material. And until now, mostly free of charge. This is now set to end.

According to a report by heise.de, the federal states have agreed on key points for a "Digital Media State Treaty". One of the key proposals: Anyone who trains or operates AI models with protected content will have to pay for it in future.

In future, there should be a licensing system that runs via collecting societies - as has long been the case with music. Authors, journalists, artists and creatives should finally have a share in the use of their works.

Protection for content - clear rules for AI

But it's not just about money. The new Interstate Media Treaty is also intended to create clear rules for the use of AI. This includes, for example, that providers must disclose what content they have used to train the models. Web crawlers, i.e. programs that automatically search the web for training material, should also be visible and labeled.

There are already certain exceptions in the EU for so-called "text and data mining" (TDM) use, for example for scientific purposes. However, many of these rules date back to a time when generative AI was still a dream of the future. This is why the countries now want to have the laws reviewed to see whether they still fit with reality.

The so-called reservation of use is also important here: rights holders - for example publishers or freelance authors - may expressly prohibit the use of their works. For this to work, however, clear and easy-to-apply standards are needed. This is exactly what the federal states now want to put in place.

Labeling AI content - strengthening real journalism

Another point in the planned treaty concerns the presentation of content on the internet. The current situation is often as follows: AI-generated content appears high up in search results, while elaborately researched articles disappear behind them - especially if they are behind paywalls.

This is set to change. Content that has been created according to journalistic standards should be more visible in future. There is also a debate about whether there should be mandatory source citations and plausibility checks for AI content. This would make it easier for users to distinguish between what is well-founded and what is simply generated text without a source.

Comment:

What is planned here was long overdue. It is unacceptable that large AI companies earn billions - with content that others have created, often with a great deal of research and time, and often without anything in return. A call for fairness is absolutely essential.

But it is also clear that these rules will only make a difference if they are consistently enforced. Tech giants in particular are often based outside the EU, are reluctant to show their cards and usually react to new regulations with a delay - or a smile.

Will the planned "AI fee" really happen in the end? Hard to say. But the fact that it is being discussed at all shows that it is: The days when AI companies could help themselves without being asked are coming to an end. And that's a good thing. Because in the digital age, we don't need less protection for creatives - we need more.

Sources: heise.de, Broadcasting Commission