Is AI displacing the creatives?

We live in turbulent times, disruption is the word of the hour. Just a short time ago, it was hard to imagine how abruptly the rules of global politics that were thought to be certain can be overridden. This also applies to the creative industries. Following the destruction of traditional distribution models for music, film, literature and art by internet platforms, a second tsunami now seems to be overrunning the creative industries: Artificial intelligence (AI) is now threatening to dismantle production models as well. How should we react to this?

Could it be that composers, writers, fashion photographers or directors will no longer be needed in the foreseeable future? Systems such as Chat-GPT, Deepseek, Claude or, especially in music, Mubert or Suno now automatically create works according to instructions formulated in everyday language: "Compose a tango in the style of Piazzolla for flute and chamber orchestra! Write a pop song with reggae elements as an ode to the mountains!" Something like this is enough as an instruction, and even musically illiterate people can have the machine cobble together pieces according to their wishes.

However, AI models only do this after the providers have trained them with real-life models. They thus use works that should actually be protected by copyright. The question at the heart of the PGM meeting chaired by National Councillor Stefan Müller-Altermatt was therefore: Are there legal options for asserting claims against the providers of AI systems?

Legal subtleties

Chantal Bolzern, a lawyer, and Noah Martin, a legal expert, both of whom have worked or are working for the Swiss rights management company Suisa, explained the difficulties involved. As their presentations showed, there is still a great deal of confusion in this area. This starts with the question of who to take legal action against and where. The players are globally active, with complex international company structures, which has already made it difficult to fight legal battles with platforms such as Facebook. It is also completely unclear how AI providers use original works to train their models and which laws are relevant: In addition to copyright law, patent and competition law could also come into play, depending on the situation.

Chantal Bolzern therefore emphasized that the key prerequisite for legal certainty in this area is transparency. We need to know exactly how the training is carried out. However, AI providers are hardly interested in disclosing this. They protect their business models with secrecy. As a result, there are debates that sometimes seem more like verbiage. The legally defined concept of "use", which allows authors to control their own works, seems to be central.

AI providers take the view that they do not "use" the original works to train their systems, but - and this is where it gets legally tricky - that they merely "enjoy" them, which can be done freely under copyright law. "Enjoyment of a work" is also a legally defined term and means "the mere taking note of a work". This, in turn, is always freely permitted. And even if there were arguments in favor of the AI providers "using" the works in the legal sense, they could retreat to the position that they are merely "using them for scientific purposes", which in turn is freely possible.

If the legal basis for dealing with the new phenomenon of AI is still completely unclear, this applies even more to the political, economic and social consequences. Ueli Schmezer, a new member of the National Council and musician who was also present at the meeting, pointed out that a motion by Balthasar Glättli on law enforcement on the internet has been pending for years. It should lead to platforms such as Facebook setting up a national contact point. Politicians have not yet made any progress here either. An effective legal strategy with a view to the upcoming challenges posed by AI therefore seems a long way off. And when you consider that politicians actually have completely different concerns at the moment, difficult times are likely to be ahead for creatives.

Constant hunger for original works

However, some schadenfreude may remain, as there are signs that the artificial intelligence that is currently sweeping across the creative industries is imploding faster than one might think. At the PGM meeting, legal representatives pointed out that the AI's statistical models always look for the average in the corpus of training data. With each new round of training, the data is also likely to include AI-generated works, making the results increasingly flat and meaningless. The systems will only retain their power if they are constantly fed with original works that deviate from the masses.

What does this mean for the business models of creatives? Faced with perplexity, it's tempting to ask Chat-GPT and its cronies. "AI models have the potential to significantly impact the business models of creative industries. They offer both opportunities and challenges as they change the creative process and create new possibilities," Chat-GPT replies. Well, we've come this far already.