French Copyright framework for artificial intelligence: a half-hearted attempt

Preamble of the draft law

The draft law begins with an explanation of the reasons for its creation. The main rationale of the proposal appears to be to “provide imperative protection for authors and artists of creation and interpretation in accordance with a humanist principle, in legal agreement with the Code of Intellectual Property”, or to “seek and find a solution to curb what already seems to us to be a threat and probably a future disaster for creation”. These alarmed comments are illustrated by the creation in 2016 of the painting “The next Rembrandt” by a computer and a 3D printer centuries after the death of the painter of the same name [more information here].

To counter this, MEPs are proposing several courses of action. These include controlling the use of works generated by AI to facilitate the collection of remuneration received for these works in order to guarantee fair and equitable remuneration for their use, ensuring the traceability of authors and artists whose work has been used by AI, and ultimately a taxation system based on collective management.

Articles of the draft law

This draft law is divided into 4 articles which amend pre-existing articles of the IPC.

The first provision amends article L 131-3 of the CPI by adding a paragraph. Under this paragraph, “the integration by AI software of intellectual works protected by copyright in its system is subject to the provisions of the CPI, as is its exploitation, and therefore to authorisation from the author or rightful owner”.

The first observation concerns the consistency of codification. This Kat questions the point of adding such a paragraph to CPI provision that deals with the formalism of copyright assignment. Does such a provision really belong here? This Kat also wishes to point out that although the intention of clarifying the current situation is commendable, the reproduction right may already cover all or part of this situation, particularly as regards exploitation and authorisation by the author. It should be noted, however, that if such an inclusion requires authorisation from the author of the original work, this can very quickly prove complicated in practice. Collective management therefore seems an interesting option [e.g., reprographic reproduction rights in France].

Article 2 adds nine paragraphs to article L 321-2 of the CPI. As a reminder, this provision concerns the ability of collecting societies to take legal action and to sit on certain bodies. The first paragraph of the new version of the article creates copyright ownership for the benefit of the authors or beneficiaries of the works used to create the artificial work, in cases where the work is created by an AI without direct human intervention. The following paragraphs aim to establish collective management of the rights generated by AI and to manage the remuneration received in this respect by collecting societies.

Once again, it is worth highlighting the inconsistency in terms of codification, particularly for paragraph 1 of article 2 of this proposal. As for the rest of the paragraphs, it is also questionable to include them in a provision which is related to collective management but which concerns, in particular, the capacity to take legal action. For the sake of clarity of the law, it seems preferable to create new articles dedicated to this collective management for works included in works created by AI.

Article 3 adds a paragraph to article L 121-2 of the CPI. Under this new paragraph, the words “work generated by AI”, as well as the names of the authors of the works that led to such a work should be added to all works generated by an AI. Once again, there is an inconsistency in the codification as the above-mentioned article of the CPI concerns the right of disclosure during the author’s lifetime and posthumously. While this new paragraph seeks to contrast the divide between AI-generated and non-AI-generated works, it has no place here. Isn’t it better to create a new article? As for inserting the name of the author of the work used, isn’t this ultimately a question of respecting authorship rights, for which provisions already exist? It should be added that these references echo the mandatory words “retouched photograph” that must accompany photographs of models used for commercial purposes and inserted in advertising messages that are broadcast. MPs seem to be following the same logic here.

Finally, the bill ends with article 4, which amends article L 121-2 of the CPI, for which the same reservations apply as regards the consistency of codification mentioned above. These three paragraphs would introduce a system of taxation borne by the company operating the AI system for the benefit of a collective management organisation for all works used whose origin cannot be determined. Article 4 notably recalls article L 324-14 of the CPI, which deals with cases where it is impossible to identify the rightholders. In that situation, the sums that should have been collected are considered as non-distributable sums, the distribution of which seems to be determined by the general policies of the collecting societies. In the draft, the law would go further by transforming these non-distributable sums into a tax intended for the promotion of creation.

In conclusion, the bill may have some difficulties achieving the goal of governing the development of AI under copyright law. It does have the merit of tackling part of the problem, namely the remuneration of authors whose work has been used by an AI to create a new work, by introducing a collective management system. Nevertheless, this bill could be criticised for being a missed opportunity. In particular, it contains many errors or inconsistencies of codification. Its relationship with the exceptions and limitations to copyright raises questions, particularly in relation to the exception for quotation, parody or pastiche. This is not mentioned in the draft law. What would happen if the AI creates a work of parody based on a protectable work ? This point would need to be clarified especially concerning the remuneration of authors.