AI Training & Usage Disclosure
This policy governs our current pre-launch submission process. The full policy set will be finalized before creator accounts open.
Everything below describes how SIREN and the Platform treat Creator content once you are onboarded as a Creator — it does not describe or apply to your current founding-creator intake submission. Material submitted through the founding-creator intake is held privately for manual review and is not processed by SIREN or used for any purpose beyond evaluating your submission — see the Acceptable Use & Community Guidelines.
1. What SIREN does
SIREN is an explain-only AI companion. It provides contextual, cultural, and analytical responses to listeners and creators about tracks, artists, and the Platform. Per architecture invariants, SIREN does not write to core records or make business decisions.
2. What SIREN processes
To answer, SIREN processes: the current track/page context, public metadata, creator-provided bio/links, lyrics (if supplied), and your prompt. It may send relevant context to third-party LLM providers to generate a response (see Subprocessor Policy / DPAs).
3. Training — confirmed position (matches the locked SIREN architecture)
- We do not use Creator content (master audio, lyrics, artwork) to train foundation models.
- We do not use SIREN to generate derivative musical or artistic works from a Creator's content.
- Any future use of Creator content for model training or generative output will require the Creator's separate, explicit, opt-in consent, presented clearly, revocable, and never bundled into general terms.
4. Listener AI interactions
Listener prompts to SIREN may be processed by LLM providers under DPAs. We do not sell prompt data. Retention per the Privacy Policy.
4a. Platform Usage Data — separate from Creator content
As we build and test new features — including Creator-facing tools — we use Platform Usage Data: interaction and engagement patterns, feature usage, session/device information, and similar telemetry, including how Creators use features while we're developing and testing them, so we can build the best possible version of each feature. Platform Usage Data does not include Creator content (master audio, lyrics, artwork) or the substance of your prompts to SIREN, which remain subject to the no-training commitment in Section 3 above — analyzing how a feature is used is not the same as analyzing or training on what you upload through it. Where technically feasible, we store Platform Usage Data in encrypted, pseudonymized, or derived form to limit its value if ever accessed without authorization; this is a security measure and does not change what data category it is. We do not use Platform Usage Data to reconstruct or approximate Creator content as a way around Section 3.
5. Providers, retention and changes
The current AI provider, processing purpose and applicable status are maintained in the versioned Subprocessor Policy. We do not state that a provider, data-retention setting, DPA or no-training configuration applies unless it is current and verified. A material change to data categories, provider use or training purpose is subject to notice and any required consent before it takes effect.
6. Accuracy
SIREN output may be imperfect or incomplete and is informational only.
7. Transparency
Where SIREN grounds a response in external search or platform knowledge, we aim to make that behavior explainable.
Reminder: the sections above describe SIREN's treatment of Creator content once you are onboarded — they do not apply to your current founding-creator intake submission.
Last updated 2026-08-27