Copyright v. artificial Intelligence: Machine learning is not a Copyright use of a Work

The article is devoted to the issue of machine learning of artificial intelligence in terms of assessing such a primarily technical process from the point of view of copyright, namely, whether it should be considered that in this case the use of works is carried out. Given that such machine learning is usually carried out at the expense of protected works, as well as given the ambiguity of the attitude to this issue in society, which, on the one hand, is manifested in high-profile lawsuits against AI developers such as The New York Times v. OpenAI & Microsoft Inc. and, on the other hand, in various social initiatives aimed at promoting the development of AI tools and deepening public understanding of the benefits of AI, the issue under consideration is extremely important and interesting.

The article focuses on the latest trends in the field of legislative initiatives with an emphasis on the law of the European Union. In particular, a brief description is given of the recently adopted Regulation (EU) 2024/1689 of June 13, 2024, which is also known as the EU Artificial Intelligence Law. It is noted that in the case of AI training using works that have passed into the public domain, the output results of AI will correspond to the level of knowledge of a century ago, i.e., will be irrelevant.

In addition, old works can only be used in machine learning if they are digitized, which can only be done manually. This will slow down the development of the latest technologies in every sense. It is substantiated that what happens during AI machine learning is nothing more than «reading»: a special form of use of a work to which the author’s right does not apply.

It is established that the use of a work for the purpose of machine learning is subject to the restriction on in-depth analysis of text and data regulated by Article 4 of Directive 2019/790/EU. It is proved that cultural creativity is unthinkable without the use of previous achievements of other authors. Therefore, since the output of AI is not a copyrightable object, in the context of AI machine learning, there are no actions significant from the point of view of copyright