Mistral's proposal to support European creation: pay for AI

Europe is approaching the issue of training AI models with fragmented and hard-to-apply rules, laments Arthur Mensch, co-founder and CEO of Mistral in an opinion piece published by the Financial Times. The current system, based on an "opt-out" mechanism that allows rights holders to refuse the use of their content, does not function properly, according to him.

A tax to get out of the deadlock

Protected works, which continue to circulate freely online, remain accessible for training models. « The legal mechanisms designed to protect them remain fragmented, inconsistently applied, and too complex », the leader estimates. Result: no one benefits from it. Creators fear a loss of revenue without a real means of control, and AI companies operate in a legal uncertainty that hinders investments in Europe.

That is why Mistral proposes a solution: the introduction of a levy on the revenues of AI companies operating in Europe. In other words... a tax, yes, it's not very original, but the French champion of AI has a plan. This mechanism would apply to all model providers, whether European or foreign, as soon as they offer their services on the European market. It is about restoring fair competition conditions and generating resources to support creation in Europe.

The funds collected would be transferred into a dedicated European fund, tasked with financing the production of new content and supporting cultural industries. In exchange, AI developers would benefit from an essential element: legal security. The mechanism would allow them to protect against potential lawsuits related to the use of online accessible content for training their models.

Arthur Mensch insists on a point, however: this contribution would not replace licensing agreements, such as the one between OpenAI and Le Monde for example. They would retain their relevance, especially for specific uses of the content. The fund would come as a complement.

There is an urgent need to act, both for creators and AI companies. The American and Chinese actors don't have this kind of problem, or rather, less. They benefit from a much more permissive framework (or even nonexistent) that allows them to train their models on massive data volumes, including European ones. The risk of the old continent falling behind is therefore real, with the prospect of becoming a mere consumer of foreign technologies.

👉🏻 Follow real-time tech news: add 01net to your Google sources, and subscribe to our WhatsApp channel.

Source : FT