AI Transparency
BASE Station voluntarily labels every sound recording in alignment with the track-level GenAI labeling program introduced by the music community (IFPI, RIAA, A2IM, WIN, IMPALA, The Grammys, SAG-AFTRA & the Human Artistry Campaign — July 2026).
A label is a disclosure, not a detection. On work made here it reflects what our own pipeline recorded; on an upload it reflects what the artist attested. We have no classifier that can listen to a finished recording and tell you how it was made, and we do not pretend otherwise.
What the badges mean
AI-Generated
Generative AI created the entirety or the primary portion of the sound recording — for example an AI-generated lead vocal, a key AI-generated instrumental performance, or fully prompt-generated music.
AI-Assisted
The recording was created substantially by humans and expresses human creativity, but generative AI was used for some expressive elements. Humans performed the lead vocal and primary instruments.
Human
No generative AI was used in the sound recording.
Per the program's guidelines, these labels apply only to generative AI used in the sound recording itself — not lyrics, composition, music videos, or cover art. BASE Station goes further and voluntarily scores those too via the Creative Ownership Score. Official announcement.
Who determines the label
Uploaded tracks: the artist self-declares the label at submission, attesting to how generative AI was used in the recording. Like the program itself, this is a voluntary, honesty-based disclosure — there is no automated "percentage of AI" measurement.
Tracks created in BASE Station studios: the label is stamped automatically by our pipeline. Fully prompt-generated music and AI vocals are labeled AI-Generated; AI harmonies and AI mastering of a human recording are labeled AI-Assisted.
Derived works (stems, masters, mashups) inherit the most AI-intensive label in their provenance chain: AI-Generated > AI-Assisted > Human.
Where origin was never declared and we observed no AI use, the label stays Unverified. That is an absence of information, not an accusation — we will not record a human performance as AI-generated because nobody filled in a form.
Provenance manifests & DDEX attribution
Beyond the track-level label, every scored asset carries a machine-readable DDEX-style AI attribution profile, exportable from its Provenance Manifest — so distributors and partner channels receive structured disclosure data instead of a single opaque label. What it carries is our record of how the work was made; it is not a certification issued by DDEX, a distributor, or any standards body.
Songwriting engines & style references
The Lyrics Studio offers three engine modes — Basic, Pro Songwriter, and 243 Masters (lyrics plus chord progressions, arrangement, and a production brief). Pro and Masters accept a reference writer or artist name.
Style, not identity: reference names are converted into abstract craft descriptors only — rhyme scheme, tempo range, prosody, narrative tone, and genre conventions. No lyrics, recordings, voice, or likeness of the referenced artist are copied or simulated, and the artist's name never appears in your output, metadata, or credits. Users may not market resulting works as being "by" or "in the voice of" a real artist (see Terms of Use §5).
Human-in-the-loop provenance: your structural choices in these engines (references, BPM, rhyme scheme, arrangement decisions) are logged as human participation signals in the work's Provenance Manifest and Creative Ownership Score.
Why we do this & reporting
We believe listeners deserve to know how the music they hear was made, and human artistry deserves clear recognition. Accurate labeling also supports the music community's push for broad, harmonized adoption of GenAI disclosure across the ecosystem — a program designed to evolve as technology and requirements change.
If you believe a track is mislabeled, contact us through the Help Center. Misrepresenting AI usage violates our Terms of Use and may result in content removal or account suspension.