Apple Music Now Labels AI-Generated Content

On its music streaming service, Apple is requesting record labels and musicians to voluntarily label tracks created with artificial intelligence. Music Business Worldwide revealed  through a newsletter to industry partners yesterday that tracks, compositions, artwork, and music videos are the four areas covered by Apple Music’s new “Transparency Tags” metadata system.

The “Transparency Tags” are a new metadata system that Apple Music has implemented to reveal the use of AI in music and visual content. These tags, which were introduced earlier this week, enable record labels and distributors to identify content in which AI was utilised for a “material portion” of the production.

When “a material portion of a sound recording” has been produced by AI techniques, the track tag should be used; other AI-generated compositional elements, including song lyrics, should be covered by the composition tag. Both static and moving visuals are covered by the artwork tag, but only at the album level. The music video tag should be used for all other AI-generated visual content, whether it is included with albums or stands alone. For works that need more than one of these disclosures, several transparency tags can be utilised concurrently.

Apple claims in its newsletter that labels and distributors “must take an active role in reporting when the content they deliver is created using AI” and that its new tags are a “concrete first step” toward attaining industry-wide transparency regarding AI-generated music.

Apple Music’s tagging system is in line with past initiatives from rival music streaming companies to shield real musicians from impersonation and spam while also making it simpler for customers to recognise AI-generated music. In collaboration with DDEX, a music standards-setting body that presently has senior Apple Music executive Nick Williamson on its board, Spotify is creating a new metadata standard for AI music disclosures. While Qobuz unveiled its own proprietary AI identification method last week, Deezer also made its AI music recognition tool from last year available to other platforms in January.

Apple Music’s Transparency Tags are completely optional (for the time being) and place the onus of AI disclosure squarely on record labels and music distributors rather than the platform, in contrast to Deezer and Qobuz’s proactive detection systems. Apple also states that content producers will have the last say over what constitutes AI-generated music and images, “similar to genres, credits, and other metadata,” and that no AI usage will be presumed on works that providers have not marked.

Other AI labelling systems have not yet been successful with honesty policies. I find it difficult to understand why record labels and creators would be encouraged to use Apple Music’s tagging system given its lax enforcement.

The important tag types are distinctively grouped into four distinct tag kinds and are used by the system to detect AI involvement:

• Track: Only available at the track level, this indicates that a significant chunk of the actual sound recording was created or altered using AI.
• Composition: Applied when AI produced a sizable amount of the song’s melody or lyrics.
• Artwork: Indicates an album or a single piece of art, both static and motion graphics, where a material component was created using artificial intelligence.
• Music Video: Recognises visual components produced by AI in standalone or packaged music videos.

In execution as well as accountability, for the time being, the system is optional for content that already exists, but sources suggest that it will eventually become a requirement for the delivery of new content. 

Self-reporting, where Apple depends on distributors and labels to manually disclose the use of AI; if a tag is missing, it is presumed that no AI is being used. 

However, in the industry context, this action is in line with Spotify’s voluntary disclosure policy, although it deviates from services like Deezer, which proactively detect machine-generated music using internal AI recognition algorithms.