How journalism can survive in the age of AI

To the point

• The value of human creativity: Josef Drexl, Director at the Max Planck Institute for Innovation and Competition, emphasizes that AI depends on human creativity and that human thinking is essential for democracy. Journalists play a key role that AI cannot replace.
• New remuneration model: The scientist proposes that AI providers pay a levy that is distributed to authors, regardless of the use of their works. This is intended to create a fair balance.
• Levy for journalism: He recommends a fee, similar to the German broadcasting fee, to promote quality in journalism and strengthen journalists as the fourth estate.

A court ruling with limited effect

In November 2025, GEMA achieved a victory against OpenAI in the Munich I Regional Court: if an AI model stores song lyrics in such a way that users can reproduce them through prompts, this constitutes copyright infringement, according to the court.

Josef Drexl makes clear that this legal victory obscures the real problem. This is because this type of reproduction through so-called memorization only occurs in relatively few cases. On the contrary, according to the logic pursued and technologically implemented by generative AI, memorization should be prevented and completely new content generated. And this is precisely where the core problem lies: AI is displacing human-created works from the market without copying them.

Substitution instead of imitation

It is important to distinguish between imitation competition (where someone copies an existing work) and substitution competition (where something new fulfills the same function). This is because copyright protects against imitation competition, while allowing substitution competition.

With generative AI, however, it is precisely substitution competition that leads to the problem. An AI article may not take a single sentence from existing texts - and yet it still displaces journalists. "Traditional copyright law falls short here structurally," argues Josef Drexl.

Why human creativity remains worth protecting

The author cites two key arguments: Generative AI is technologically dependent on human creativity. Models that are trained too extensively with AI-generated data will increasingly "hallucinate," i.e., produce more and more errors and eventually even collapse.

Even more important is that our democracy needs human thinking. Journalists, as the so-called fourth estate, not only inform the public, but also uncover grievance through their research, thereby keeping those in power in check. Creative professionals in the cultural industry play a similarly important role. "Generative AI alone cannot do this," emphasizes Josef Drexl.

The radical proposal: Participation without reference to the work

Drexl outlines three levels of regulation - from optimizations of the existing system to a new "right to AI exploitation" to his preferred model: a "right to fair compensation," completely detached from the specific use of the work.

Josef Drexl proposes that commercial providers and operators of generative AI pay levies that collecting societies distribute to authors - as a non-transferable right. The distribution would take the form of a surcharge on other distributions over the last five years. AI developers should also be able to accept this, because "the development of high-quality generative AI models is centrally dependent on being trained with human-generated products," according to Drexl.

The advantages are obvious: there are neither problems of proof nor lengthy negotiations. Creative professionals whose works have never been used for AI training also deserve compensation, because they too are exposed to competition from AI-generated output.

Thinking beyond copyright

Drexl emphasizes that copyright alone cannot solve the problem. For journalism, he proposes a citizen's levy, similar to the broadcasting license fee in Germany. Voters would decide annually which press publishers receive the funds, provided that quality standards are met and a minimum percentage of publishing costs is based on the remuneration of journalists. This would create a disincentive to rationalize with generative AI.