Suno: A Music Copyright Minefield

The emergence of AI-generated music has created an alarming ease with which streaming platforms can be inundated with unauthorized imitations of popular artists, such as Beyoncé.

While AI music platform Suno explicitly states its policy prohibits the use of copyrighted material, allowing users to remix their own tracks or set original lyrics to AI-generated music, its intended copyright recognition and blocking mechanisms for existing songs and lyrics are proving remarkably easy to circumvent. Although no system is infallible, Suno's filters have demonstrated significant vulnerabilities.

Demonstrating minimal effort and leveraging readily available free software, users can prompt Suno to produce AI-generated renditions of well-known tracks, including Beyoncé’s “Freedom,” Black Sabbath’s “Paranoid,” and Aqua’s “Barbie Girl.” These imitations are strikingly similar to their originals; while most listeners might discern a difference, some could plausibly mistake them for alternative versions or B-sides upon casual listening. Furthermore, there is a tangible risk that these "uncanny valley" covers could be exported and uploaded to streaming services, leading to their monetization. Suno declined to provide a comment regarding these findings.

The creation of these unauthorized covers necessitates access to Suno Studio, a feature included in the company’s $24-a-month Premier Plan. Unlike text-based song generation, Suno Studio facilitates the uploading of existing tracks for editing or covering. While the platform is generally designed to detect and reject unmodified, well-known hits, simple manipulations using free tools like Audacity can bypass these safeguards. For instance, altering a track’s speed to half or double its original tempo, or adding a brief burst of white noise at its beginning and end, often proves effective in circumventing the filter. Users can then revert the speed and remove the white noise within Suno Studio, effectively using the copyrighted song as the foundation for new AI-generated music.

When generating a cover from imported audio without specific style transfers, Suno's models exhibit varying degrees of adherence to the original. Models 4.5 or 4.5+ largely reproduce the instrumental arrangement with only minor alterations to the sound palette. In contrast, Model v5 takes greater creative liberties with the source material, exemplified by its addition of "chugging guitar and galloping piano" to “Freedom” and its transformation of the Dead Kennedys’ “California Über Alles” into a "fiddle-driven jig."

Suno also provides functionality for adding vocals, either by generating lyrics or allowing users to input text. While the system is designed to block copyrighted lyrical content, direct copying and pasting of official lyrics from sources like Genius will result in the system flagging them and producing unintelligible vocals. However, even minimal alterations to these lyrics can successfully bypass this protective filter.

During testing, it was possible to deceive Suno Studio by making minor spelling adjustments to a few words within “Freedom” – for instance, altering “rain on this bitter love” to “reign on” and “tell the sweet I’m new” to “tell the suite.” Beyond the initial verse and chorus, such modifications became unnecessary. The resulting AI-generated vocals closely mimicked the original recordings, creating slightly distorted but recognizable impersonations of artists like Ozzy Osbourne or Beyoncé.

Independent artists may face even less protection. During a test of the company’s v5 model, one of my personal songs bypassed the copyright filter entirely. Similarly, tracks by singer-songwriter Matt Wilson, Charles Bissell’s “Car Colors,” and experimental artist Claire Rousay successfully cleared Suno’s copyright detection system without any modifications whatsoever. This suggests that artists operating under smaller labels or distributing independently via platforms like Bandcamp or services such as DistroKid are particularly susceptible to having their work copied. DistroKid and CD Baby both declined to comment on this issue.

The output of these AI-generated covers often resides squarely within the "uncanny valley." While the source material remains unmistakable—the riff from “Paranoid” is instantly recognizable, and “Freedom” is clearly “Freedom” from its opening snare hit—the renditions possess a distinct lack of vitality. Even an alarmingly accurate-sounding AI version of Ozzy Osbourne, for instance, exhibits a deficit in nuance and dynamic range, conveying the impression of a human imitation rather than an authentic performance.

Instrumental components in these AI covers similarly either omit the original’s distinctive artistic choices or replicate them as bland imitations. For example, a non-jig cover of “California Über Alles” loses its characteristic raw edges, sounding akin to a generic wedding band rendition. Pink Floyd’s “Another Brick in the Wall,” originally an experimental "doom disco" track, is reduced to vacuous dancefloor filler. While the AI might approximate David Gilmour’s guitar tone, it sacrifices any sense of phrasing or progression, rendering the solo a mere "mindless stream of notes."

The creation of these unauthorized covers directly contravenes both Suno’s stated mission and its terms of service. Furthermore, Suno’s copyright scanning appears to be limited to the initial upload phase; there is no apparent re-checking of generated outputs for potential infringement, nor are tracks rescanned prior to export. This oversight creates a straightforward path for monetizing Suno-created covers. Individuals could easily upload these renditions via distribution services such as DistroKid, thereby profiting from others’ intellectual property without remitting the standard royalties typically associated with cover songs for the original composers. Independent artists, in particular, appear to be the most susceptible to this exploitation.

Folk artist Murphy Campbell recently encountered this issue firsthand when apparent AI covers of her YouTube-posted songs were uploaded to her Spotify profile. (The specific generation system used remains unclear.) Subsequently, the distributor Vydia initiated copyright claims against Campbell’s original YouTube videos and began collecting royalties. Highlighting the systemic flaws, the very songs for which Vydia successfully claimed copyright were, in fact, in the public domain. While Spotify eventually removed the AI covers and Vydia rescinded its claims, this resolution only occurred after a concerted social media campaign by Campbell. Vydia maintains that the two incidents are separate and denies any association with the AI covers of Campbell’s work.

The issue of AI-generated fakes extends to numerous other artists. Experimental composer William Basinski and indie rock group King Gizzard and The Lizard Wizard, for instance, have both experienced imitations of their work bypassing multiple filters and appearing on streaming platforms such as Spotify. In some instances, these counterfeit tracks can even divert listens directly from the legitimate artist’s page. Within a streaming ecosystem where artist payouts are often meager—Spotify, for example, mandates a minimum of 1,000 streams for payment eligibility—less prominent musicians bear the brunt of these challenges.

Suno, therefore, represents just one component within a demonstrably flawed system.

Streaming services including Deezer, Qobuz, and Spotify have implemented various measures to counteract "spammy AI" content and impersonators. Chris Macowski, a spokesperson for Spotify, informed The Verge that the company “takes protecting artists’ rights seriously, and approaches it from multiple angles. That includes safeguards to help prevent unauthorized content from being uploaded in the first place, along with systems that can identify duplicate or highly similar tracks. Those systems are backed by human review to make sure we’re getting it right.” Despite these efforts, no system is entirely flawless, and managing the deluge of AI-generated content facilitated by platforms like Suno presents an ongoing and significant challenge.

Macowski further acknowledged the inherent technical complexities, stating, “It’s an area we’re continuing to invest in and evolve, especially as new technologies emerge.”

Ultimately, Suno is but a single element within an unequivocally broken system, yet it represents a point against which artists have notably limited recourse. While bands can appeal to Spotify to remove AI-generated fakes from their profiles, ascertaining the origin of these fakes—and specifically whether they stem from Suno’s filter failures—proves challenging. To date, Suno itself has maintained silence on the matter.