TuneCore flags Papa Michigan track as AI despite human-recording appeal

TuneCore flagged a song by dancehall veteran Papa Michigan as apparently made with unlicensed generative-AI tools, despite the artist’s assertion that it was recorded entirely by people in a studio. The disputed track was Track 7, “Who Can Be Against Me,” from the album So Many Roads. The Jamaica Gleaner reported that TuneCore told Papa Michigan the song was ineligible for distribution because it had identified indicators of generative-AI tools trained on improperly licensed datasets. Papa Michigan responded on social media by showing the studio and musicians involved in the recording. He described it as a “100-per-cent human studio recording,” said AI had not been used at any stage, and lodged a formal appeal seeking review by a human moderator. He later removed the track and released So Many Roads on Aug. 14. The album is available across major streaming platforms through Rashanco Music. Evon Mullings, general manager of the Jamaica Music Society, said false positives involving human-made music are becoming increasingly common. He said distributors including TuneCore, DistroKid and CD Baby face pressure from services such as Apple Music, Spotify and Deezer to screen AI-generated and AI-assisted music before it reaches retail platforms, leading to aggressive automated detection. Mullings said artists can challenge a decision through a distributor’s support team by providing proof of creation. Useful material can include studio video, high-resolution digital-audio-workstation screenshots and raw audio stems. He said highly quantised drums or heavily processed, aggressively Auto-Tuned vocals can make a human recording appear robotic to an automated system. Separately, Believe and TuneCore have deployed detection technology which Believe chief executive Denis Ladegaillerie described as 99% reliable and capable of identifying the AI model or platform involved, Music Business Worldwide reported. TuneCore blocks releases when it identifies an unlicensed service. The policy illustrates the tension facing distributors: detection systems are meant to prevent AI music created with unlicensed services from entering streaming catalogues, but an erroneous flag can stop a human artist’s release. Mullings said streaming services generally seek disclosure and labelling rather than an outright ban on AI-made music. He said Apple Music and YouTube rely on uploader disclosures, Deezer uses in-house detection to tag fully AI-generated tracks, and Spotify relies on voluntary disclosure and metadata tags. The push for labelling is also tied to consumer transparency, competitive concerns and the distribution of streaming royalties. Mullings said standard streaming royalties are allocated according to a track’s share of total plays, leaving human artists vulnerable when bot-driven streams of AI tracks draw from the royalty pool. He said many services exclude AI-made music from editorial playlists, algorithmic recommendations and official charts. JAMMS is introducing revised song-registration requirements to reflect emerging global practices for music rights in the AI era.