With the agreement on the AI Act, Europe has been described as a pioneer in the regulation of artificial intelligence. In the European debate, Vincent Lorphelin denounces the excessive focus on the risks of piracy for authors, which blurs the real issues at stake.
The IA Act, the world's first legislation on artificial intelligence, is presented as a landmark agreement. Yet it is undermined by a constellation of explosive postures from two opposing camps: pro-innovation versus pro-culture.
In the first camp, Jean-Noël Barrot, Minister for the Digital Economy, intends to gut the agreement during the forthcoming discussions on its implementation. He is backed by Emmanuel Macron, who wants to protect the French unicorn Mistral AI. This ChatGPT competitor is a major European hope, threatened with death by the AI Act. All its value lies in its secret of refining the Internet to feed generative AI, which the AI Act would force it to reveal. Unsurprisingly, the Committee for Generative AI, launched by Elisabeth Borne, over-represents innovators.
In the opposite camp, authors consider that their works have been massively plundered, which represents an existential risk. They are demanding transparency on the content used to drive generative AI and substantial compensation. They are supported by MEPs, who defended the controversial clause in the AI Act, and indirectly by European Commissioner Thierry Breton, who criticizes those who are concerned with private interests rather than the general interest.
An absurd situation
The radicalization of these positions has led to an absurd situation. First of all, while the threats are indeed serious on both sides, they rest primarily on the legal assessment of whether or not works are fair use. Final decisions may not be made for another ten years. In the meantime, this uncertainty will hamper investment far more than the AI Act. Secondly, it is technically doubtful that the transparency imposed by the AI Act is necessary for authors to enforce their rights. Finally, the French government finds itself defending innovation against culture, which contradicts its historic vocation for the cultural exception.
This situation can only be detrimental to France and Europe. In fact, it reveals a deep-seated unthinking about the desirable balance between human and artificial intelligence.
Just as we have long considered the authoring professions to be the last bastion of human intelligence, the breakthrough of generative AI is seen by Elon Musk and other followers of the Singularity Theory as proof of its forthcoming domination. However, this breakthrough only reveals that even these professions have their share of repetitiveness. Generative AI does not replace, but remains a tool at the service of creative activity. This is the conclusion of the agreement reached between screenwriters and Hollywood studios after a six-month strike. It was also the Copyright Office's recent decision that the production of a generative AI cannot be protected by copyright, as it does not express the personality of an author.
Allocating royalties
This legal clarification must now be matched by an economic balance. We have seen how the platform model has imposed itself at the expense of Internet contributors. Even when countries such as Australia and Canada throw their weight behind the balance of power, they can only extract a pitiful 2% royalty to remunerate their press. Current negotiations with media groups Gannett and News Corp show that this rate will be crushed by generative AI.
To stave off this doom, the New York Times is taking legal action against Open AI and Microsoft to revalue its rights. Major image databases, such as Shutterstock, are launching their own generative AIs to compete with the leaders while remunerating copyright holders. This movement heralds a new wave of generative AI, in the same way that iTunes and Spotify innovated to overcome mass music piracy.
The current political debate is focused on the modern version of the Hadopi law, which was intended to curb piracy, while the next wave is already arriving. All players need to review their positions. Collective management organizations need to embrace generative AI. Operators of generative AI must renounce sources that infringe copyright. Instead, they must provide indicators to fairly distribute royalties to authors according to their contributions. The government must accompany negotiations to secure their legal risks. Finally, it must think about the desirable balance between intelligences, to anticipate the crushing of royalties. The cultural exception, which has led discotheques to pay musicians and Netflix to finance cinema, must inspire it to set the overall rate to be paid by generative AIs.