Sony’s new tech can sniff out original songs in plagiarized AI-generated music

Neural watermarking is a technology that embeds digital markers into copyrighted music, allowing detection systems to identify when original songs have been used as training data or incorporated into AI-generated compositions. Unlike surface-level audio comparison that matches frequencies or waveforms, neural watermarking works by analyzing the deeper musical structure—including melodic patterns, chord progressions, and harmonic elements—to identify similarities even when songs are altered through changes in tempo, key, arrangement, or remixing. This enables platforms to trace the origin of musical elements and determine whether a song contains portions derived from copyrighted works, helping artists recognize their influence on AI tracks and establish grounds for potential compensation.

Identifying original songs within AI-generated music is crucial because AI music generation systems have been trained on vast collections of copyrighted music without permission, raising significant copyright infringement concerns. When AI systems generate new compositions, they may incorporate melodic fragments, chord sequences, or harmonic elements from their training data, creating derivative works that infringe on artists' intellectual property rights. By detecting these original songs embedded in AI-generated music, copyright holders can establish evidence of infringement, pursue legal action for compensation, and ensure that artists are properly credited and compensated for their work. This technology also helps distinguish between coincidental musical similarity and actual copying, providing nuance in determining whether infringement has occurred.