I’m doing some research around publishing in France, particularly in regards to copyrightability of partially AI-assisted or wholly AI-generated works. Like many things these days, I began my inquiry with good old CG, who turned me on to this article on Mondaq from Dec. 2023, with its own interpretation of the text:

One of the key proposals includes amendments to the French Intellectual Property Code (IPC). These amendments suggest that when an AI system creates a work from unidentified sources, a tax should be imposed on the AI system’s operator. This tax would be directed to copyright collective management organizations, which typically represent authors and other right holders in the management of their rights. However, the allocation of this tax revenue to the actual creators of the works used by the AI remains unclear, and it seems that the benefits of this tax might not directly reach the creators

ChatGPT

This assessment seems to be mostly borne out by the Mondaq version, though I don’t know quite how reliable their hot take is either. Perhaps somewhat, according to Wikipedia?

Highlights from the Mondaq itself:

Thise [sic] article imposes tax on the operators of AI systems, wherein the work generated by AI is based on works whose origin cannot be identified. The net effect of such addition would be taxing entities like Google, Microsoft and OpenAI to name a few who are required to pay the French collection societies. […], since the proposed Article only applies to works whose origin cannot be determined, it falls short in creating any apparent benefit for the authors or owners, as effectively, the benefits of this article will never be experienced by creators since the revenues collected will form part of non-distributable sums, which will be distributed as per the general policies of such collection societies.

TechDirt seems to confirm the above with their own interpretation here. Article 4 of this proposal seems to be the origin of the tax supposedly to be paid by AI companies to collective/collecting societies, which may be effectively impossible to actually pay out ot individual creators? If that’s an accurate interpretation, it doesn’t sound very great for anybody potentially but the societies. Techdirt quotes extensively from this article on the legislation by Andres Guadamuz, dated Sept. 2024. Quoting:

As you can see from the analysis of Article 2, allocating an author may be either impractical or impossible. So in my view, this is the real intention of the legislation. As it will be not possible to find the author of an AI work (which remember, has copyright and therefore isn’t in the public domain), the law will place a tax on the company that operates the service… the tax will be paid by OpenAI, Google, Midjourney, StabilityAI, etc. But also by any open source operator and other AI providers (Hugginface, etc). And the tax will be used to fund the collective societies in France… so unless people are willing to join these societies from abroad, they will get nothing, and these bodies will reap the rewards.

[…] what will likely happen is that all of the companies will cease to provide services in France, much in line with what happened in Canada with Canadian media, and French AI creators will turn to open source models in the wild with no curator or institution guarding it, and these will not pay any tax. That, or VPNs will become very popular.

Which only strengthens my impression that we have to be careful and circumspect in how we draft potential new legislation to fix what we think the “harms” of AI are or might be. Because we might end up making the situation worse for the very people we think we were supposed to be helping originally…

I’m still investigating this, but based on dates in the above articles, I am guessing none of this has passed yet, but worth keeping an eye on for sure!