We don’t train on your data
Your content is used to answer your visitors’ questions — never to train AI models. The AI providers we rely on are contractually barred from training on it too.
Trust & security
Your data stays yours.
Your content is there to answer your visitors’ questions — nothing else. Here’s exactly where it lives, who can touch it, and how it’s handled.
We don’t train AI on your content. Full stop. Your documents, your visitors’ questions, and the conversations they have with your bot are used for one thing: answering those questions. We never use them to train models, and the AI providers we build on are contractually prohibited from training on them either. Your content stays your content.
How your data is protected
The short, checkable version.
Stored in the EU
Your content and conversations are stored at rest in the European Union. AI processing happens under the strict no-training, short-retention terms described below.
Encrypted, audited infrastructure
Data is encrypted in transit (TLS) and at rest (AES-256), on infrastructure that is independently audited to SOC 2 Type II and ISO 27001.
Isolated per customer
Every bot’s content and conversations are walled off from every other customer at the database level — enforced by row-level security, not just application code.
Deletion that deletes
Remove a source and its text and the search index built from it are erased together, in one operation — so removed content can no longer be retrieved or cited.
Minimal third-party processing
To generate an answer, your text is sent to an AI provider, processed, and automatically deleted within 30 days. It is never retained to train their models.
Where your data goes
Who processes it, and why.
- Cloud platform (EU) — Hosts the database, authentication, and the application. All data encrypted at rest; independently audited.
- AI inference providers — Generate answers and the search embeddings from your content. Operate under commercial terms that prohibit training on your data and auto-delete it within 30 days.
- Email delivery — Sends handoff and escalation notifications to your team. Receives only what’s needed to deliver the message.
This is the role-level view. A complete, named subprocessor list is provided in our Data Processing Agreement (DPA), available on request.
Private & internal bots
Internal data is a different conversation.
A bot over your public help center and a bot over sensitive internal knowledge aren’t the same risk, and they shouldn’t be set up the same way.
For private or internal knowledge bases we offer dedicated, single-tenant isolation, stricter retention, and processing kept inside your own environment. Those are bespoke deployments — tell us what you need and we’ll scope it with you.
Running a security review?
We can provide our Data Processing Agreement, the full named subprocessor list, and answers to your security questionnaire. Ask when you request access — we read every message.