French CSPLA Report Highlights AI Regulation: Transparency and Copyright Compliance in Focus

The CSPLA, an advisory body to the French Ministry of Culture, was entrusted with a mission to clarify the scope of the transparency obligation and to propose a summary template that could be supported by France at the EU level. The mission, led by Professor Alexandra Bensamoun, consulted various stakeholders, including authors, right holders, AI model providers, and public authorities. The mission delivered a report that focused on the field of copyright and related rights, and did not address other issues such as personal data protection, competition law, or cultural diversity.

Analysis

The CSPLA report provides a comprehensive and balanced analysis of the challenges and opportunities posed by the use of data and content for AI training, from the perspective of copyright and related rights. The report acknowledges the importance of data and content as strategic inputs for AI innovation, but also stresses the need to ensure the respect of the rights and interests of authors, performers, and other right holders, as well as the values and rights of the EU. The report proposes a pragmatic and proportionate approach to the transparency obligation, based on the following principles:

• The transparency obligation consists of two inseparable obligations: to establish a policy to comply with EU copyright law, and to disclose a sufficiently detailed summary of the content used for AI training. The summary should include relevant elements of the compliance policy, especially those related to the opt-out clause provided by the 2019 directive on copyright and related rights in the digital single market.
• The transparency obligation does not imply self-regulation by the AI model providers, and may require the disclosure of a list of the content used, depending on the reliability of the sources. The summary should not be limited to listing the main data sources, but should provide enough information to enable the right holders to identify the potential use of their content and to exercise and enforce their rights. However, the summary should not reveal technical information that could harm the trade secrets of the AI model providers, such as the filtering or tokenization processes. The summary should only disclose the ingredients, not the recipe, of the AI model. 
• The transparency obligation creates a bridge to the respect of copyright and related rights, but does not provide a specific procedure for this matter. The report suggests two possible ways to facilitate the exchange of information and the resolution of disputes between the right holders and the AI model providers, without prejudice to judicial actions: a direct dialogue, facilitated by a contact point indicated in the summary, and a complaint procedure, involving the European AI Office or a national authority, with the possibility of mediation. In both cases, the AI model providers should be able to prove that they comply with the law and that they did not use any content without authorization.

The CSPLA report has important implications for the industry and the public, as it sets out clear and reasonable expectations for the AI model providers regarding the transparency of data use and the respect of copyright and related rights. The report also contributes to the development of a trustworthy and human-centric AI framework in the EU, which could enhance the competitiveness and the innovation potential of the European actors in this field. The report is likely to have an impact at the EU level, as it provides valuable insights and recommendations for the elaboration of the summary template by the European AI Office, and for the interpretation and application of the regulation by the EU institutions and the member states.