In a striking statement during an interview, Microsoft’s big AI boss takes a surprising position on online copyright: He doesn’t seem to really believe in its existence.

Mustafa Suleyman, CEO of Microsoft’s AI division, does not believe in copyright on the Internet. This is evident from a very striking passage in a recent interview. In it, Suleyman refers to a self-invented social contract that would trump copyright law worldwide.

Sophisticated partnership agreement

Suleyman: “I think that when it comes to content that is already available on the open web, the social contract of that content since the 1990s has been that it is.” fair use Is. Anyone can copy it, recreate it, reproduce it. That is Freeware If you want, that’s the deal.”

To be clear: this is an objective lie. Just because you can see, record or copy something does not mean you can do whatever you want with it online or offline. Belgian copyright regulations therefore meet the international standard. The Ministry of Justice states literally on its website that copyright also applies on the Internet. In other words: copying and reproduction is completely prohibited without the author’s consent.

Suleyman’s comments illustrate how the tech giants view publicly available data. In any case, the CEO of Microsoft AI insists that it can use whatever is useful without regard to the authors. This position is diametrically opposed to that of the authors themselves.

Trained in disability work

Finally, the manufacturers of large AI models such as OpenAI are currently under criticism for their handling of data from the Internet. Large language models (LLMs), on which generative AI applications are based, have been trained with huge amounts of data from the Internet.

For example, ChatGPT can generate intelligent responses thanks to New York Times articles, and Getty Images sued Stable Diffusion because the model learned to draw from millions of Getty images. More recently, record labels have joined forces to sue AI music generators. If you ask Suno or Udio to compose a country or rock song, they will succeed because the models have processed tens of thousands of country and rock songs during their training.

We are now seeing the first licensing agreements between AI companies and organizations like Reddit that manage and own a lot of content. This way, AI companies hope to avoid future lawsuits.

Inadequate legislation

Suleyman is completely wrong with his invention fair usepolicy, but there is a problem with the current legislation. After all, this mainly relates to the reproduction of protected works, but says nothing about the training of AI models. Copying a painting is not allowed, but what if you use a hundred paintings as a basis to create a new similar painting? Is that a reproduction? Or is it about something else? This question is currently the subject of an ongoing legal dispute.