SACEM's European federation calls for trustworthy AI records

The audio task

A composer registered with SACEM, the French authors' and composers' society, wants to know whether a generative AI system can legally train on their catalogue and, if so, whether anyone will ever be able to prove it happened. That second question, about proof, is what drove a joint statement from SACEM's European federation of collecting societies during the EU's AI Act negotiations.

What the documents show

SACEM is a listed member society of GESAC, the European Grouping of Societies of Authors and Composers, which represents 32 national CMOs. In a statement dated Brussels, 20 October 2023, issued ahead of an AI Act trilogue meeting, GESAC called record-keeping and transparency obligations on generative AI models' use of copyrighted content 'necessary to ensure responsible and trustworthy AI systems.' The statement asked EU legislators to require that generative AI models demonstrate compliance with the opt-out mechanism in the EU's text-and-data-mining exception, and to make AI output detectable as artificially generated. Separately, SACEM's own 'Sacem and Artificial Intelligence' notice, published 27 May 2025, lists the questions its members are asking, including whether AI tools can use their works without authorisation, and points members to a separate position document rather than answering those questions itself on the page.

Rights status

Neither document states that SACEM signed an individually named 'trustworthy AI' charter. What is verifiable is that GESAC, SACEM's European federation, publicly asked EU institutions for specific transparency and record-keeping rules in the AI Act, describing them as the basis for trustworthy AI systems, and that this was advocacy directed at legislators still negotiating the text, not a description of enacted law. SACEM's own 2025 notice states it has a position document but frames the underlying questions, including registration and remuneration, as open rather than settled.

What to check before you use it

This is an editorial checklist. Anyone relying on 'SACEM supports trustworthy AI' as a specific commercial commitment should check GESAC's own statement text rather than a paraphrase, since it addresses EU legislative process, not a licence SACEM offers directly. A rights holder should also check whether the AI Act's general-purpose AI obligations, as finally adopted, actually match what GESAC asked for in October 2023, since advocacy and outcome are not the same document.

• Did the AI Act's final text include the record-keeping obligation GESAC asked for in October 2023?
• Does SACEM's own position document, referenced but not fully reproduced on its news page, add anything the GESAC statement does not?
• Is an AI vendor's claim of 'trustworthy AI' backed by disclosed training records, or only by marketing language?

The traceable fact here is a federation-level advocacy statement naming record-keeping as the mechanism for trustworthy AI, not a signed charter with an AI company, and the two should not be conflated when summarising SACEM's position.