Privacy & Data Practices
Last updated: April 4, 2026 · Aiboxz / Nova Pro
This page summarizes how we handle customer data for the Nova voice terminal and related services. It is written for operators who must answer security questionnaires without waiting for legal counsel on every line item.
Data isolation
Session expert memory and persona-scoped Redis streams are keyed by a stable customer identifier derived from your authenticated account (not by ephemeral room names). Custom personas and memory partitions are not commingled across tenants at the application layer.
Private RAG readiness
Retrieval blocks injected into the voice agent are sourced from your own session memory and optional knowledge bases you control. We do not operate a global cross-customer embedding pool for Nova session recall.
No training on user data
We do not use your voice transcripts, expert memory payloads, or exported research cards to train foundation models. Inference is performed through standard provider APIs under terms that exclude your content from model training where commercially available.
Contact
For data processing questions: legal@aiboxz.com (placeholder — replace with your operational inbox).
This document is informational and does not replace a signed Data Processing Agreement for Enterprise customers.