Data and privacy questions
Where your data lives, what the model sees, what is never stored, and how long records are kept.
Where your data lives
The platform runs in the EU (Frankfurt region) by default, and your business data stays inside your tenant boundary — tenant isolation is enforced as row-level security in the database on every table. Per-client region pinning is committed before general availability.
Is my data used to train models?
No. NeoXam never fine-tunes models on your data, and there is no learning loop that feeds your business data back into any model. Model providers are used for inference only, accessed through an AI gateway — and switching providers remains a configuration choice precisely because nothing is trained on your data.
What the model sees
For each run, the model receives the agent's compiled system prompt, your input, and the tool results needed for that bounded task — nothing more. Credentials never appear in prompts, logs, or traces: a redaction layer enforces this, with automated contract tests guaranteeing that no prompt or completion content leaks into telemetry.
Retention
The audit log is append-only and designed for 7-year retention — the regulatory minimum for the vertical — configurable per tenant. Conversational session history (Beta) is archived and reviewable.
Note
During the Beta, trace storage may use a managed observability service under a disclosed, opt-in scope — never for client business data — and moves fully inside the deployment boundary before GA.
Need more?
For deployment options and the full trust posture, see the Security page under Platform — or contact us. We answer security questionnaires as part of the Early Adopter Program.
Still stuck?
Tell us what you were trying to do and where it failed — screenshots and run IDs help. A human follows up.