Built to pass your safety & procurement review.
This is the layer a rigorous reviewer cares about. The short version: it's language infrastructure, not patient data — and it was sourced the right way.
Not patient data. Not PHI.
BTS-BH100 describes how a language names and talks about the body and health — terms, grammar, phrasing, voice. It contains no patients, no records, no identifiable health information. It sits outside HIPAA/PHI scope: it's the linguistic substrate a clinical model needs, not the clinical data itself.
Consent-first — including voice
Every contribution, text and audio, is given voluntarily under explicit consent, with grievance and withdrawal channels. Voice recordings are scoped and de-identified by design.
Fairly sourced
8 native contributors, fairly compensated. This is sourced, not scraped — which is also why it can't simply be reproduced from the open web.
Evaluate before you commit
Labs get scoped, de-identified access to evaluate the substrate directly on their own infrastructure — data access only, no black box. A pilot opens every domain, including this one, before any full license.