Privacy Tier

Session audio is the most sensitive piece of data a clinical system processes. The two 2026 AEPD (Spain's data protection authority) notes on AI transcription say it plainly: voice is personal data in the strict sense whenever it can identify the client, and the minimisation principle requires keeping it only for the time strictly necessary for the purpose of treatment. But "the time strictly necessary" varies by clinical case — and that is why nexmin does not impose a single policy. STANDARD is the default and meets GDPR out of the box: audio comes in, is transcribed, normalised for future analyses and the original is deleted. Only the lightweight processed file remains, so the AI can revisit it later if you need to re-analyse the session with a new prompt or a different model. LEGACY keeps the raw audio in cold storage when you have given the client an explicit commitment to be able to return their "full session" on demand — useful in some deeper therapeutic contexts where the client wants that record preserved. GHOST is the strictest option: after the next analysis, the processed file is also deleted and only the transcript and the clinical synthesis remain. The decision is yours, not the system's. When you create the client's record you choose the tier; you can change it later, and the change applies to material captured from that moment onward (not retroactively, so as not to destroy what the client authorised when they signed the original consent).

Related terms

• Identity Vault
• Data sovereignty