The creative paradox: Generative AI and the copyright crisis

Artificial Intelligence has leaped from being a simple analytical tool to becoming a creator in its own right. Platforms that generate coherent texts, photorealistic images, or original musical scores are now accessible to everyone, giving rise to the phenomenon of Generative AI. But how do these machines manage to “create” and—more importantly—what are the ethical and legal implications of this revolution?

Generative AI operates through complex models, such as Generative Adversarial Networks (GANs) or Transformers, which are trained on enormous datasets of pre-existing works: millions of texts, images, or audio tracks. It is crucial to understand that these systems do not simply copy the works; rather, they learn the style, composition, and relationships between elements in the training data. Once these creative “rules” are learned, the algorithm can generate completely new content in response to a simple text command, the so-called prompt.

The training mechanism is the real Gordian knot causing a crisis in the world of copyright. To be effective, AI requires a colossal amount of data, and most of this comes from publicly accessible works on the web, often protected by copyright. When AI generates a work, it is the result of a “statistical average” of what it has learned from millions of sources. The crucial question facing the legal world is this: Is the use of protected works to train Generative AI a copyright infringement?

The Paradox: Picasso, the Human Assimilator, and the Algorithm

This is the paradox that captures attention. A good way to understand this dilemma is through a historical parallel with a great artist: Pablo Picasso.

Picasso did not invent Cubism out of nothing. His works are deeply rooted in the study, assimilation, and re-elaboration of thousands of stimuli: from African masks to Iberian art, from Cézanne’s techniques to the masterpieces of the classical masters. Picasso learned from all these sources to create something radically new, and no one would question the originality and authorship of his works.

Similarly, Generative AI imitates this learning process on an industrial scale. The AI “looks” at millions of images, recognizes the patterns, and finally, upon user command, generates an output. The fundamental difference, which is the crux of the legal and ethical dispute, is that the AI operates without consciousness, without intent, and on a mass of data often not explicitly authorized for commercial use. While Picasso is a human “transformative assimilator,” the AI is only an algorithmic “transformative assimilator.” This distinction is crucial and opens deep wounds in copyright law.

The Legal Implications of Generative AI

Courts worldwide are beginning to address this unprecedented challenge, which focuses on two key points. The first concerns the Originality of the Output: Can the work created by Generative AI be considered original and thus protected by copyright? If the law requires a “spark of human creativity,” to what extent can a user’s prompt confer originality on an algorithm-produced work? The second point is Dataset Infringement: Artists and publishing companies argue that the mere act of copying and analyzing their works for training constitutes an infringement, even if the final result is a different work.

Drive2Data‘ Approach

In the face of these complexities, we at Drive2Data approach the concept of Generative AI with a pragmatic and ethical vision. We recognize that AI is a disruptive tool, but we emphasize its nature as an accelerator and not a substitute for human creativity. The focus is on data governance: it is not just about producing content with AI, but about ensuring that the datasets used are compliant and that companies fully understand the legal and ethical risks arising from an unconscious use of models trained on unverified data. Essentially, for us, the innovation of Generative AI must always go hand in hand with responsibility and transparency.

In conclusion, while the learning and re-elaboration process of Generative AI might statistically mirror that of a great master like Picasso, its nature remains non-human, and the sheer volume of training data opens up a huge regulatory void. The goal will be to find a sustainable balance between protecting the rights of creators and the incredible thrust for innovation offered by this technology, regardless of the profound meaning of “artistic creation.” This is the ontological side of art itself and the seemingly indissoluble knot that perhaps holds the solution to the entire dilemma.

MORE TO EXPLORE…