Suno’s AI Music Raises Copyright Concerns
▼ Summary
– Suno’s platform policy prohibits using copyrighted material, but its filters are easily bypassed with simple audio edits like speed changes or adding white noise.
– Users can generate AI covers of popular songs that closely mimic the originals, which could then be monetized on streaming services without paying royalties.
– The platform’s lyric filter for vocals can also be tricked with minor textual changes, such as tweaking a few words in a song’s lyrics.
– Independent and lesser-known artists are particularly vulnerable, as their original tracks sometimes pass through Suno’s copyright detection without any alterations.
– While streaming services have measures to combat unauthorized AI content, the system remains flawed, and Suno has not publicly addressed these filter failures.
While Suno’s official policy explicitly prohibits the use of copyrighted material, its safeguards are proving to be surprisingly porous. The AI music platform, which allows users to upload their own tracks for remixing or to set original lyrics to AI-generated music, is designed to block the use of protected songs and lyrics. However, its copyright filters are easily bypassed with simple, widely available techniques, raising significant concerns for artists and rights holders.
Using Suno Studio, a feature on the company’s $24 monthly Premier Plan, individuals can upload audio to create covers or edits. The system will typically reject an unaltered, well-known hit. Yet, by employing free software like Audacity to perform basic manipulations, such as changing a track’s speed or adding short bursts of white noise, users can consistently circumvent the detection system. Once the modified file is uploaded, the original speed can be restored and the noise removed, effectively using the copyrighted song as a direct blueprint for new AI-generated music.
The resulting tracks often land in an uncanny valley of audio. When generating a cover without applying a style transfer, especially using Suno’s model 4.5 or 4.5+, the output can be an instrumental arrangement nearly identical to the original, with only minor tonal adjustments. Model v5 takes more liberties, but the core melody and structure remain unmistakable. Tests produced recognizable AI imitations of hits like Beyoncé’s “Freedom” and Black Sabbath’s “Paranoid.” The vocal generation feature, which is also supposed to block copyrighted lyrics, can be tricked with minor textual alterations, like changing a few words in a song’s verses.
Perhaps more troubling is that independent artists receive even less protection. During testing, original songs by lesser-known musicians passed through Suno’s copyright filter without any modifications at all. This suggests that artists on small labels or who self-distribute are at the highest risk of having their work co-opted without detection. The financial implications are direct and damaging. Once a convincing AI cover is generated in Suno, it can be exported and uploaded to streaming platforms via a distribution service. This allows bad actors to potentially monetize these imitations, diverting royalties away from the original creators without paying the standard licensing fees for a legal cover.
The consequences are not theoretical. Folk artist Murphy Campbell recently found AI covers of her YouTube-posted songs uploaded to her official Spotify profile. A distributor then filed copyright claims against her own videos, attempting to collect royalties, despite the songs in question being in the public domain. The situation was only resolved after a public campaign. Other artists, including William Basinski and King Gizzard and The Lizard Wizard, have also battled unauthorized AI impersonations on streaming platforms. For musicians who already operate on razor-thin margins in a streaming economy, such infringements can have a tangible impact on their livelihood.
A critical flaw in the process is that Suno appears to scan for copyright only upon upload, not re-checking the final AI output or the exported file for infringement. This creates a clear pathway for abuse. While streaming services like Spotify state they employ safeguards and human review to combat spam and impersonations, they acknowledge the escalating challenge posed by the volume of AI-generated content. As one Spotify spokesperson noted, it is a complex technical area requiring continuous investment.
The silence from Suno in response to these findings is notable. The platform represents a single component in a broader ecosystem that is struggling to adapt, but its vulnerable detection systems place a disproportionate burden on artists, particularly independent ones, to police and protect their own work. The ease with which its filters are defeated highlights a growing gap between policy and practical enforcement in the age of generative AI.
(Source: The Verge)
