ACTA Brief 01 — The Evidence Gap: When AI Classification Determines Royalties, Visibility and Eligibility, What Evidence Does an Artist Have?
Between June and September 2026 the recorded-music industry adopted, in rapid succession, a set of instruments for handling AI-generated and AI-assisted music: automated detection with economic consequences (TIDAL, July 15), industry-wide voluntary labels (IFPI, RIAA and partners, July 10), platform-side disclosure and identity badges (Spotify AI Credits, AI Persona from mid-September), mandatory machine-readable marking of synthetic audio in the European Union (AI Act, Article 50, applicable August 2), and a patent licensing framework for the generative pipeline (Music IP Holdings, August 20). Each instrument produces a classification with consequences: royalty eligibility, visibility in recommendations, badges, in some cases removal. Several of them now provide an appeal mechanism for misclassification. None of them specifies what evidence an artist can bring to such an appeal, and none standardises an artist-held, creation-time evidentiary record against which such a classification can be checked. This brief names that gap, maps the layers of the current stack by what each one establishes, presents a prospective case (the false positive), and locates a creation-time, artist-held evidence layer within the existing infrastructure rather than against it. It makes no claim that such a layer determines authorship or that a work is human-made. It argues that classification without evidence is now an operational problem with a price, and that the price is being paid first by the parties that pay artists.