Spotify Announces New AI Safeguards, Says It's Removed 75M 'Spammy' Tracks (variety.com)

Spotify announced tightened defenses against AI-driven fraud and “spammy” music, saying it removed more than 75 million tracks in the past year and is rolling out new policies and systems to curb deepfakes, mass uploads and other manipulative practices. Key measures include a stricter impersonation policy that bars unauthorized AI voice clones unless the artist consents, faster content-mismatch review and pre-release reporting for artists, and a new music spam filter (launching this fall) that will tag uploaders and tracks engaging in mass uploads, duplicates, SEO hacks and artificially short tracks — then prevent those items from being recommended. Spotify says it will deploy the filter conservatively and keep adding detection signals as new schemes emerge. Spotify is also backing an industry-wide disclosure standard through DDEX so creators and rightsholders can indicate where and how AI was used (vocals, instrumentation, post-production), with metadata pushed from labels and distributors and shown in-app. For the AI/ML community this matters because it raises the bar for provenance, attribution and detection tools: platforms will need robust signal engineering and forensic models to separate legitimate creative use from content farms, while distributors will be locked into interoperable metadata practices. The move aims to protect artist royalties and trust in streaming, but also signals an accelerating arms race between generative tools and platform-level defenses.