AI detection benchmark reveals music artifact patterns and limits
MUSEBENCH: A Benchmark for AI-Generated Music Detection and Artifact Analysis
Sound
Summary
It is becoming very hard to tell AI-generated music from human-made music. The authors created MUSEBENCH, a huge collection of music tracks made by humans and AI, to test how well detection tools work. They analyzed 11 detectors using many types of AI music and tricks that change the sounds to see where detectors succeed or fail. They also studied typical AI music glitches, like strange chords and unrealistic sounds. This research helps improve tools for spotting AI-created music in the future.
What this means in practice
- •For music streaming platforms: Automatically detect AI-generated music uploads to enforce content policies and maintain catalog authenticity.
- •For digital rights managers: Identify AI contributions in musical works to aid copyright management and royalty distribution.
Authors
Yi Chen Liu, Syed Irfan Ali Meerza, Jiawei Yu, Xi Guo, Samuel Bright, Jian Liu
Abstract
Advances in generative AI have made it possible to produce music that is nearly indistinguishable from human compositions, raising the question of whether such works can be reliably distinguished from human compositions. We present MUSEBENCH, a large-scale benchmark for AI-generated music detection and artifact analysis, comprising over 361K tracks. The dataset includes human compositions, stylistically matched music generated by ten state-of-the-art open-source and commercial models, such as Suno and Udio, as well as human-AI co-creations and variants processed with adversarial perturbations, compression, and acoustic post-processing. Using MUSEBENCH, we benchmark 11 automated detectors and evaluate their accuracy, generalization, and robustness, under default, out-of-domain, and transformed-variant experiment settings. We further conduct a comprehensive artifact analysis through acoustic feature inspection, identifying recurring signatures such as non-idiomatic harmonic progressions, upsampling induced spectral periodicities, and violations of physical performance constraints. Our findings reveal key strengths and vulnerabilities of existing detectors and provide practical insights for building more reliable AI music detection systems.