Google announced it will begin labeling AI-generated songs in YouTube Music, rolling out the feature this month. This marks the company's first major update to its AI content labeling policy since earlier this year when it started adding labels to AI-generated videos on YouTube.
The new labels will be visible on the player page, and users can tap to view more information. The labels will be applied to songs that were created entirely or significantly altered with AI. However, the exact method for identifying AI-generated music remains unclear, as Google is relying on both voluntary disclosures from publishers and its own detection tools.
"Generative AI is opening up new ways for artists to create, experiment, and share music," said Google in a blog post. The company emphasized that it is not banning AI music but is instead providing transparency for users. This approach is similar to the AI labeling system Google introduced for YouTube videos earlier this year.
The labeling process will involve three tiers: fully AI-generated, partially AI-generated, and no AI involvement. Publishers can choose to disclose how much AI was used, but this is optional. Google also mentioned using SynthID watermarking and C2PA metadata to detect AI-generated content.
While Google is better positioned than most streaming services to identify AI content, the absence of a label does not guarantee a track is human-made. This is true for all AI content labeling. The company acknowledged that AI tools can be used in ways that leave no identifying mark in the final file.
Google did not say how users can block labeled AI tracks from appearing in their playlists, and it remains unclear how effective the labels will be for music compared to YouTube videos. AI labeling for music is now rolling out in YouTube Music, with AI-generated YouTube videos set to receive the same treatment soon.
Source: arstechnica
