Music streaming services—like virtually every other corner of the internet—are awash in AI slop, thanks to the rise of cheaply accessible AI tools and a regulatory void. Now, a coalition of organizations representing musicians and record labels is trying to make it easier for listeners to at least know when a song has been generated by an LLM.
Led by the Recording Industry Association of America (RIAA) and the International Federation of the Phonographic Industry, the group is pushing streaming platforms like Spotify and Apple Music to adopt a voluntary labeling system for identifying songs that were made using AI, according to a Friday report from the Wall Street Journal.
The proposed labels would appear to listeners similarly to the label currently used to mark songs containing explicit content. They look a bit like tiles from the periodic table: One of them, designed for tracks generated entirely by AI, has a black background with “AI” written in big, white letters; another, made for songs created by humans who used AI for specific parts of the creative process, has a white background with “ai” in smaller letters. At this point, the coalition is not suggesting the adoption of labels that would mark AI-generated cover art or music videos, according to the WSJ report.
The effort—which is being backed by entertainment industry giants like the Grammys and SAG-Aftra—is intended to add a measure of transparency around AI-generated music, not to suppress it entirely.
It’s the latest in a long history of organizations trying to come to terms with the rampant rise of online deepfakes. While European companies are required under the EU’s AI Act to ensure that AI-generated “detectable as artificially generated or manipulated,” no such governmental regulation currently exists in the U.S. That means the decision to label AI slop—or not—falls on the shoulders of American tech companies themselves. Some social media platforms, like Instagram and TikTok, can automatically apply labels to posts in which AI-generated content is detected. Elon Musk’s X is infamously less stringent on this front, though that platform uses a “community notes” system, designed to allow users to call out AI slop if and when they see it. The late, controversial Sora app, which OpenAI shut down in April, also automatically applied a label to its AI-generated videos, though users quickly figured out creative ways to remove it.
AI-generated music is notoriously harder to spot than AI-generated images. The latter includes static metadata that can be read by AI-detection algorithms relatively easily and is often addled with telltale markers, like extra fingers and warped letters. The metadata for AI-generated music, in contrast, is dynamic and fluctuating, making it easier to evade human or algorithmic detection. In 2024, a song featuring AI voice clones of Drake and The Weeknd blew up online, with millions of listeners mistaking it for an actual, entirely human-made track.
The viral potential of AI-generated music was clearly too good a market opportunity for Spotify to pass up. Earlier this year, the streaming giant introduced a new feature enabling premium users to generate AI remixes of songs from participating artists. The feature, launched as part of Spotify’s new partnership with Universal Music Group, was positioned as a means of empowering artists in the age of AI, giving them a chance to profit from the new generative AI wave rather than sitting idly by and letting it slowly eat up their streaming revenue.
Spotify announced an “impersonation policy” in September, mandating that any songs featuring AI voice clones of an actual artist be published with the consent of that artist. The platform’s app also offers a voluntary feature through which artists can list the use of AI in a song’s credits. Similarly, Apple Music recently introduced optional “transparency tags” for music or artwork on the platform made with the help of AI.
Read the full article here
