AI Policy

1. Our position

MGWP Music is an independent label. The music in our catalogue is human-made. We use software — including algorithmic and AI-assisted tools — to run and present the label, and this page says exactly where and how. Transparency is not a disclaimer for us; it is part of the product.

2. What this platform uses

• Visualizers (EQ strip & party ring) — deterministic signal processing of the live audio (waveform + BPM + genre). No generative model, no playback-history adaptation.
• Cover text — rendered programmatically from each release's own metadata, so a cover can never contradict its track.
• Recommendations & playlists — a rule-based scoring engine over catalogue data (BPM, genre, play counts). No profiling beyond the catalogue.
• Operations — automated QA (metadata checks, cover OCR audits, test suites). No AI reproduces or imitates artist voices or music.

3. AI Credits — per release

Every release declares the tools used in its production. The default is none: no generative AI in the music, vocals, or artwork. This is published mechanically: each release page carries an “AI transparency” block (audio / cover art / lyrics / curation) sourced from the release's own metadata, with the values mirrored in machine-readable data-ai-* attributes for agents and crawlers. Where a release does use AI-assisted elements, its declaration says so by name.

Artists on the label can request the disclosure template at [email protected].

3b. Provenance records

Each release also publishes a provenance record: a manifest binding the AI-usage declaration above to the release's identity, its track references and a SHA-256 content hash. It is served as JSON at /api/releases/<id>/provenance and linked from every release page. The manifest carries a verified flag computed by re-hashing its own contents, so a hand-edited declaration stops verifying.

What it is not. It is not an official C2PA content credential, and it does not establish authorship, ownership or legal provenance.

4. Training & licensing

Our catalogue — recordings, artwork, and metadata — is not licensed for generative-AI training, fine-tuning, or inclusion in synthetic-output datasets. Access does not grant consent. We license only through an explicit, written opt-in; reach out at [email protected].

Our robots.txt and llms.txt describe machine access to public pages. Agents and assistants are welcome to read the catalogue the way a visitor would — as users, not as training corpora.

5. Human-made standard

Releases in which every creative element is human-made carry the label's human-made note. It is the default state of the catalogue, and we intend to keep it that way.

6. Contact

Licensing & AI questions: [email protected] · Copyright / takedowns: see our DMCA policy.