What legislation for generative AI?

At least since ancient Greece, the fantasy of an anthropomorphic robot with extensive learning capabilities has regularly fired the imagination, especially the literary imagination, particularly since the advent of science fiction literature. But the latest dizzying development in artificial intelligence technology does not currently present itself in the form of androids, but as data centers in which specific programs weave their artificial neural networks and evolve with the help of data available for free on the web, including conversations on social networks and texts used by online translation tools. This vast amount of "training data" inspires and influences AI programs and enables them to generate new content through algorithms based on the knowledge they acquire. Conversation tools are currently the best-known example. In response to this massive use of text, a considerable number of websites, including those of major newspapers, can now only be read for a fee. On the one hand, this contributes to limiting the offer for individual users who do not want to subscribe to multiple websites and, on the other hand, to distorting the quality of information available on the net: disinformation or ideologically oriented websites have no interest in preventing the use of their content, on the contrary, they gradually become more important when reliable news sites become inaccessible to AI search engines.

The use of AI is not limited to writing more or less elaborate texts: it also collects image, audio and video files. The voices of actors or singers are used without their consent, images of political figures are altered and video montages make them say things they never said. Aside from this worrying use, which is due more to malice and a lack of legislation than to the technology itself, the use of AI can of course make certain activities much easier and make tedious repetitive work unnecessary. This is the case, for example, with software that makes it possible to identify unwanted noises in a recording and automatically remove them, such as the crackling when digitizing an old record. Similarly, several functions of music notation software work with AI. Effectively, most people have been using this technology for several years without realizing it. The film industry uses it massively, so much so that many jobs are at risk, as the 2023 strike by American screenwriters spectacularly demonstrated. The writers concerned demanded that AI tools not be used to put them out of work, but only as a research aid or for details of the script. Actors who specialize in voice dubbing are also worried and fear being replaced by digital tools. In fact, there is already software that can clone voices and make them say anything you want. One platform already offers to create your own episodes of series and share them on the web, whose entire conception (script, text, editing, characters, dubbing, music, etc.) is carried out by AI. Even if the initial results are not yet one hundred percent satisfactory, we are at the beginning of a technological revolution that will be accessible to everyone, for better or for worse.

In addition to the issue of the protection of personal data, which is close to the hearts of internet users, the issue of copyright and related rights is the subject of ongoing attention from artists' associations. A heated debate is raging between those who are concerned about the rights of authors and seek the most appropriate jurisdiction possible, and those for whom works are merely data that can be exploited without restriction by AI. The key question will be which interests are considered paramount by legislators and the courts. To this end, the International Federation of Musicians (FIM) has drawn up a declaration with specific recommendations. The SMV is actively involved in the FIM through its Central Secretary Beat Santschi, who, we would like to remind you, is also Vice President of the Federation.

The FIM declaration

 After a brief reminder of the historical changes that the music sector has faced, FIM emphasizes that the development of the legal framework can partly help to compensate for their negative effects. Although the Rome Convention and the WIPO Performances and Phonograms Treaty (WPPT) of the World Intellectual Property Organization (WIPO) have provided welcome solutions for performers in relation to broadcasting and communication to the public, these instruments have unfortunately failed to effectively regulate downloading and streaming, as Article 10 of the WPPT in its current version does not allow performers to receive a fair share of the revenue from the online exploitation of their recordings. This legal framework is even less able to address the particular problems posed by generative AI today, whether in the use of existing data or the end result of its use.

The FIM recommendations include the following considerations: "It is unacceptable that performers can become victims of large-scale exploitation of their works, sounds, voices, images, likenesses or styles without their free, prior and informed consent and without financial compensation. Performers should have the right to authorize and effectively prohibit the scraping [automatic extraction and storage of data from websites] and analysis of their works, sounds, voices, images, likenesses or styles by an AI system, even after the transfer of their exclusive rights, and to receive financial compensation for such use. It is also necessary to ensure that performers enjoy the same level of protection against the unauthorized use of their performances by AI, regardless of whether they are based on a literary or artistic work, an expression of folklore or AI-generated material."

Financial compensation

"Once AI has recorded and analyzed the works, sounds, voices, images, likenesses or styles of artists, it can use this data to produce new content on a scale that represents a significant market distortion and an objective threat to the careers and livelihoods of all current and future artists. We need a sustainable legal and economic environment that effectively prevents AI-generated audio and visual material from distorting the market with prices that are far below those of human creations protected by copyright and related rights. To the extent that AI-generated content derives its value from human creations that are widely used, it is entirely appropriate to consider mandatory compensation mechanisms that benefit the creative community and apply to all AI generative tools. Therefore, innovative compensation mechanisms based on production should be considered. For any AI-powered generation of music content, performers should be fairly compensated, as their work and talent form the knowledge base for this content. However, such fair payments must not lead to the normalization or undue promotion of individual people's work through generative AI. We need a compensation system that honestly forces a producer considering the use of generative AI to weigh the economic benefits of products and performances created by humans against the benefits of products created by generative AI."

This topic will continue to be a source of controversy and concern in the years and decades to come. Among the interesting reflections on this topic was the symposium "Artificial Intelligence and Creativity" held in Bern in May 2024, organized by the Swiss Coalition for Cultural Diversity, Suisseculture and the Federal Film Commission. Videos of the contributions are on the Suisseculture website to find.

FIM statement on AI in music