Suno's AI music platform copyright filters can be defeated using Audacity speed changes and white noise, enabling near-identical covers of major artists.
Investigative testing revealed that Suno's copyright detection — meant to block unauthorized use of protected songs — is trivially bypassed by slowing or speeding up a track in free software like Audacity and adding white noise bookends. Suno Studio's $24/month Premier Plan allows audio uploads, and manipulated tracks of songs like Beyoncé's 'Freedom' and Black Sabbath's 'Paranoid' passed the filter, producing AI covers close enough to be mistaken for alternate takes. The resulting audio can be exported and uploaded to streaming services. Suno declined to comment when contacted.
This exposes a fundamental weakness in audio fingerprinting systems: simple time-stretch transformations destroy spectral fingerprint matches without meaningfully changing the listening experience. If you're building any audio ingestion pipeline that relies on services like ACRCloud or AudD for copyright detection, this is a live demonstration that those checks are insufficient against trivial preprocessing. The fix — perceptual hashing combined with ML-based similarity scoring — exists but costs more and has higher false-positive rates.
Run your audio moderation stack against a time-stretched sample this week: take any 30-second clip, apply 0.5x speed in FFmpeg, and submit it to your current fingerprinting endpoint to verify whether your system catches it.
Install ffmpeg and run: ffmpeg -i input.mp3 -filter:a 'atempo=0.5' output_slow.mp3
Tags
Signals by role
Also today
Tools mentioned