DataSecurity

Yourdatastaysinyourdatabase.

Anylytics brings the AI to your data, not your data to the AI. Queries run at the source. Agents work from metadata.

Data flow

What moves, and what stays put.

  1. You ask

    The question goes to the Copilot with the metadata of the scope you picked.

  2. The model writes SQL

    It sees table and column names, types, descriptions and join rules. No table contents.

  3. Anylytics checks it

    SQL outside the scope is rejected. A row limit, aliases and ordering are applied.

  4. Your database runs it

    The query executes at the source. Only the result rows come back to explain the answer.

File uploads (CSV, Excel) are the exception: they are stored and queried in the Anylytics cloud on AWS.

The model reads names, types and descriptions. Your tables stay where they are, behind checks you control.

SQL guardrails

Checks run before any SQL reaches your data.

Generated SQL goes through security levels. The strictest, Full, applies all four checks.

SQL security levels and their checks
LevelChecks
FullScope validation, row limit, column aliases, stable ordering
BasicScope validation, row limit
SQL editScope validation, for SQL you edit by hand
PaginateRow limit, for paged results
Access

Control who sees what.

Roles and privileges

Organisations, groups and roles decide who can connect sources, build scopes and share results.

Asset permissions

Every connection, scope, report and canvas carries admin, write or read access.

Secure sign-in

Sign in with your Google account. Every API call checks a signed token.

Scoped AI

Agents work inside a Data Scope. Tables outside it are invisible to the model.

Encrypted traffic

Traffic between your browser, Anylytics and the AI gateway is encrypted in transit.

AI providers

Model calls go to the providers named in our Privacy policy.

Questions about security?

Book a session with our team. We walk through the data flow on your stack.