Photo Credit: Nahrizul Kadri
One AI audio avalanche later, the RIAA, A2IM, IMPALA, and several others have introduced a labeling program designed to identify machine-generated tracks on DSPs.
The mentioned organizations, besides the IFPI, WIN, the Recording Academy, and SAG-AFTRA, introduced that straightforward program today. At the top level, the entities have proposed tagging recordings as “AI-Generated” and “AI-Assisted” moving forward.
Those tags would take the form of two suggested square-shaped icons: One, for straight artificial intelligence generations, prominently featuring “AI” in all caps, with the other, for AI-assisted recordings, positioning the lowercase “ai” towards the square’s bottom.
As many will recognize, there’s sure to be a bit of debate about precisely what constitutes an AI-assisted recording. In the organizations’ own words, audio will fall into this category if it “was created substantially by humans and expresses human creativity; however, generative AI was used for some expressive elements.”
Additionally, humans must have “performed the lead vocal and primary instruments” for a track to receive the AI-Assisted, not AI-Generated, label.
Regarding the use of “recordings,” the RIAA and others made clear that their “system does not cover the use of generative AI in lyrics, composition, music videos or cover art at this point.”
And with the labels expected to become “available for use in the near future,” all eyes are on the all-important implementation process. Unsurprisingly, given the high-profile organizations behind the push, talks are said to be underway “with digital music services, distributors, aggregators and standard-setting bodies on industry-wide implementation.”
Even so, this leads to perhaps the most pressing obstacle on the road to widespread adoption across DSPs: The outlined system revolves around “voluntary” AI-use disclosures as opposed to automatic AI tags like those found on Deezer.
In the long term, there will presumably be an opportunity to flip the switch to “mandatory”; the labels are “designed to evolve as technology and requirements change,” per the organizations.
Furthermore, it’s possible that AI tags of any kind will have the desired effect; anti-AI fans will know which songs to avoid and will perhaps gain the option of filtering machine-generated slop down the line.
Nevertheless, it doesn’t seem like a stretch to assume that artists and especially “uploaders” will be less than forthright about their use of AI. If Suno CEO Mikey Shulman’s “Ozempic of the music industry comments” are accurate, plenty of commercially prominent releases were actually pumped out with the assistance of gen AI. As such, we should be able to draw worthwhile conclusions if a curiously small number of projects are tagged accordingly.
Bigger picture, the campaign appears to demonstrate across-the-board concerns with DSPs’ AI-identification policies amid accelerating machine-generated uploads. Spotify added verified artist badges in late April, and as tracked by DMN Pro, several AI slop artist pages have experienced conspicuous monthly listener slips on the year.
But as noted, Deezer is alone in identifying and tagging AI audio from the get-go. In a statement provided to DMN, the platform expressed a willingness “to support the development of an industry-wide framework.”
“It’s encouraging to see steps being taken towards a unified approach to generative AI in music. As the first music streaming platform to detect, tag and exclude AI-generated music from algorithmic recommendations, Deezer is ready to support the development of an industry-wide framework,” Deezer said.
“This includes key considerations around the use of training data for AI models, ensuring that all rights holders are fairly remunerated. We’re looking forward to continuing our collaboration with the wider music ecosystem to create fair and practical standards for AI in music,” the DSP concluded.