The policy of Suno, an AI music platform, is that it does not allow the use of copyrighted material. You can upload your own tracks to be remixed or set your original songs to AI-generated music. But, its purpose is to prevent you from recognizing and using other people’s songs and lyrics. Now, no system is perfect, but it turns out that Listen’s copyright filter is incredibly easy to fool.
With minimal effort and some free software, Listen will offer AI-generated imitations of popular songs like Beyoncé’s “Freedom”, Black Sabbath’s “Paranoid” and Aqua’s “Barbie Girl” that are worryingly close to the originals. Most people will probably be able to tell the difference, but some may mistake them for an alternate take or B-side upon casual listening. Furthermore, it is possible that one can monetize these unique Wally Covers by exporting them and uploading them on streaming services. Suno declined to comment for this story.
Creating these covers requires using Suno Studio, which is available on the company’s $24 per month Premier plan. Instead of rendering the entire song with text, Listen Studio lets you upload a track to edit or cover. This is likely to capture and reject a well-known hit without any changes. But using a basic free tool like Audacity to slow a track down to half the speed or speed it up to twice the normal speed will often bypass the filter, and adding a blast of white noise at the beginning and end basically guarantees success. You can restore the original speed and reduce white noise in Listen Studio, and the copyrighted song becomes the seed of new AI music.
If you create a cover of imported audio without any style transfer, Listen seamlessly brings out the original instrumental arrangement with very little change to the sound palette if you’re using models 4.5 or 4.5+. Model v5 is a bit more aggressive in taking liberties with the source material, adding strumming guitars and galloping piano to “Freedom” and turning the Dead Kennedys’ “California Uber Else” into a fiddle-driven jig.
Listen lets you add vocals by composing lyrics or typing words into a box, and once again, it lets you block anything that’s copyrighted. If you copy and paste the official lyrics of a Genius song, Suno will flag them and remove the ambiguous lyrics. But even very minor changes can bypass this filter.
I was able to trick Listen Studios by changing the spelling of a handful of words in “Freedom” – “rain on this bitter love” to “reign” and “tell sweet I’m new” to “tell the suit” – and beyond the first verse and chorus, I didn’t even have to do that. The sound closely mimics the original recording, invoking a slightly off-brand rendition of Ozzy or Beyoncé.
Indie artists can’t even be given that level of protection. One of my own songs cleared the copyright filter when I was testing v5 of the company’s model. I was also able to get tracks from singer-songwriter Matt Wilson, Charles Bissell’s “Car Colors,” and experimental artist Claire Russe through Listen’s copyright detection system without any changes. Artists on smaller labels or self-distributing through services like Bandcamp or DistroKid are most likely to slip up; DistroKid and CD Baby declined to comment.
The results of these AI covers fall firmly into the uncanny valley. The songs they’re covering are unmistakable: the riff of “Paranoid” is recognizable and “Freedom” is clearly “Freedom” the moment the marching snare hits begin. But there is a lifelessness in them. Even though AI Ozzy seems dangerously accurate, it lacks nuance and dynamics, making it feel more like an imitation of a human rather than the real thing.
The instrumentals similarly discard any interesting artistic choices made by the originals, or clone them into flat imitations. A non-jig “California Uber Else” cover has had most of the rough edges sanded down, so it sounds like a wedding band version of the original. Pink Floyd’s “Another Brick in the Wall” goes from an experiment in doom disco to just empty dancefloor filler. And, while it is like David Gilmour’s guitar tone, it removes any phrasing or sense of progression, turning the solo into a mindless stream of notes.
Creating unauthorized covers violates both Suno’s stated purpose and terms of service. Additionally, Listen appears to only scan tracks upon upload; It seems it doesn’t double-check the output for potential violations, or re-scan the tracks before exporting them. From there the path to monetizing Suno-produced covers is simple. AI slopmongers can upload them through a distribution service like DistroKid and make profits from other people’s songs without paying the usual royalties a cover would pay the original musician. And independent artists appear to be the most vulnerable.
folk artist murphy campbell This was discovered recently when someone uploaded what appears to be AI covers She posted songs on her Spotify profile on YouTube. (It is unclear through what system they were generated.) Shortly afterward, distributor Viadia filed copyright claims against her YouTube videos and began collecting royalties on them. And to highlight just how broken the whole system is, the songs for which Vydia successfully filed copyright claims are all in the public domain. Spotify eventually removed AI covers, and Vydia dropped its copyright claims, but only after Campbell’s social media campaign. Vidia says the two incidents are separate and not connected to the AI ​​cover of Campbell’s work.
AI fakes are a problem for other artists as well. There are imitations of experimental musician William Basinski and indie rock group King Gizzard and The Lizard Wizard. Slip through multiple filters And access streaming platforms like Spotify. Sometimes, these fake songs can garner views directly from the artist’s page. In a system where payouts can already be extremely low – Spotify requires a minimum of 1,000 streams to get paid – lesser-known musicians suffer the most.
Listen that’s clearly just one part of a broken system.
Services like Deezer, Qobuz and Spotify have taken measures to combat spammy AI and impersonators. Spotify spokesperson Chris Makowski told The Verge The company “takes protecting artists’ rights seriously, and looks at it from multiple angles. This includes safeguards to help prevent unauthorized content from being uploaded in the first place, as well as systems that can identify duplicate or highly similar tracks. Those systems are backed by human review to make sure we’re doing it right.” But no system is perfect, and keeping up with the flood of AI slop enabled by platforms like Suno is a challenge.
Makovsky acknowledged the technical difficulties involved, saying, “This is an area we are continuing to invest in and develop, especially as new technologies emerge.”
Listen that’s clearly just one part of a broken system. But this is an artist who has particularly little recourse against which to fight. Bands can contact Spotify and have the AI ​​fakes removed from their profiles. It is hard to tell how those fakes arise, and whether they are the result of Suno’s filters failing. And so far, Suno’s reaction is silence.
