What happened

On 16 May 2024 Sony Music Group published a Declaration of AI Training Opt Out, a public statement that its affiliates Sony Music Entertainment and Sony Music Publishing reserve their catalogues from being used to train AI systems. The declaration followed reporting that record companies were sending opt-out letters to AI developers; Sony's version is the company's own public notice of that position rather than a leaked or reported letter.

What the documents say

The declaration states that Sony's affiliates "expressly prohibit and opt out of any text or data mining, web scraping or similar reproductions, extractions or uses" of their content "for any purposes, including in relation to training, developing or commercializing any AI system," except where specifically authorised. It applies to "musical compositions, lyrics, audio recordings, audiovisual recordings, artwork, images, data" and covers existing and future content, including works identifiable through databases maintained by IFPI and the International Confederation of Music Publishers. The declaration is a company statement, not a court filing or statute, so it establishes Sony's asserted position and scope, not a court's finding that the position is enforceable in every jurisdiction. Its legal footing in the EU comes from a separate document, Article 4(3) of the 2019 Copyright in the Digital Single Market Directive, which lets the EU's text-and-data-mining exception be defeated where rights are "expressly reserved by their rightholders in an appropriate manner, such as machine-readable means." A public declaration in prose is a starting point for that reservation, but the Directive itself names machine-readable means as its model example, which a webpage alone may not satisfy for every use.

Why it matters for makers

An opt-out declaration only does legal work where a jurisdiction actually recognises an opt-out mechanism, such as the EU's text-and-data-mining exception; it does not by itself amend US copyright law, where no equivalent statutory opt-out appears in these sources. For an artist signed to Sony, the declaration is the label asserting a position on the artist's behalf across the whole catalogue, not a mechanism the artist activates individually.

What to check before you use it

If you are independent and want a comparable reservation, check what format a given jurisdiction actually requires; in the EU, prose alone may be weaker evidence than a machine-readable signal a scraper's process would detect. If you are signed to a major label, this declaration means the label, not you, controls the opt-out position for your catalogue unless your contract says otherwise.

  • Does the jurisdiction relevant to my catalogue recognise an opt-out from AI training at all, and in what form?
  • If my label has made a declaration like this one, does my contract let me set my own position separately?
  • Is my own reservation, if I have one, in a machine-readable format rather than only a webpage statement?

A declaration like this asserts a position; whether it is honoured depends on a provider's compliance process and, ultimately, on enforcement, neither of which this document alone can demonstrate.

Continue in Rights & release, or EU law now makes AI model makers disclose training summaries.

Evidence & further reading

Go to the source.

Source records preserved from the earlier research edition. This pass has not independently reverified every archive claim or current policy.

  1. Declaration of AI Training Opt Out ↗

    Sony Music's own declaration of its opt-out scope and the content categories it claims to cover.

    Source date 2024-05-16 · Accessed 2026-09-16
  2. Directive (EU) 2019/790 on Copyright in the Digital Single Market ↗

    Article 4(3) text establishing the EU's machine-readable rights-reservation mechanism referenced by the opt-out.

    Source date 2019-04-17 · Accessed 2026-09-16