The model plans. It executes nothing.
The AI model holds no credentials and runs no queries. A separate execution layer validates every step, enforces read-only access, and records the whole run so you can check it.
Boundaries enforced in the architecture
Review the controls alongside the source permissions, data handling and remaining limits below.
- 01
The model proposes; services execute.
The model returns a structured plan. Backend services validate it and run the permitted steps.
- 02
Every query is validated before execution.
SQL is parsed into a syntax tree and checked against read-only rules, including statements inside CTEs. KQL and ES|QL use dedicated validators.
- 03
Analysis code runs without internet access.
Generated Python runs in an isolated sandbox. Outbound traffic is blocked at the network layer, with execution-time and memory limits. The sandbox receives query results without database credentials.
- 04
Credentials stay out of model context.
Credentials are encrypted at rest and used by backend connection services for connection tests, schema discovery and query execution.
- 05
Saved work is private to its account.
Access checks and database policies scope connections, conversations, results and dashboards to the account that created them.
- 06
Every run is recorded, step by step.
Reopen the question, structured plan, executed query, returned results and execution details from history. Analysts can inspect the calculation; report readers can check the sources and period.
The guide to read-only AI analytics: architecture, controls, and threat model explains the controls and remaining questions for a security review.
What flows where
- 01
Question
The model receives the question, schema metadata, descriptions and conversation context. Selected values from single-row results, including text, and database errors can also enter context.
- 02
Plan
The model proposes the query and analysis steps. The backend validates the plan before execution.
- 03
Execution
Connection services execute the validated read. Generated Python processes provided results in the restricted analysis sandbox.
- 04
Result
Datarelix stores returned results and run artifacts so you can view and reopen them. Multirow result tables are not sent through the single-row follow-up context mechanism.
For the full data-flow walkthrough — what the model receives, who holds the connection, and where secrets live — see how Datarelix keeps database credentials out of the AI model.
Stored information and access controls
Stored
- Account details such as name, email and sign-in method; connection configuration such as host, dialect and allowed scope.
- Encrypted connection credentials.
- Questions, schema information, plans, executed queries and returned results.
- Saved charts, dashboard definitions and service usage records.
Execution and access restrictions
- Database credentials are not supplied to the model.
- Generated reporting queries are validated for read-only execution.
- The generated Python sandbox has network egress blocked.
- The privacy policy describes service providers and the handling of personal data.
The service runs on Google Cloud. Traffic is encrypted in transit, storage is encrypted at rest, and connection credentials carry an additional layer of application-level encryption.
Authentication and access
Sign-in
Google, Microsoft (single- or multi-tenant, with optional tenant allowlisting), or email and password with verification and reset.
Your permissions, per user
On Azure SQL, Databricks, and Kusto, connections can run with the signed-in user’s own database permissions (per-user On-Behalf-Of) instead of a shared service account.
Least privilege, recommended
Connect with a read-only database user scoped to the schemas you want to ask about. Read-only is enforced at the query layer either way — a scoped user makes it defense in depth.
Questions, reviews, and reports
Security questionnaires, SOC 2 status, and DPAs go through the contact page. Vulnerability reports go straight to support@datarelix.ai. If you are running an evaluation, start from the AI analytics security checklist, written by Datarelix.
See the guardrails for yourself.
Connect a database, ask a question, and read the query that ran.
14 days free · no credit card required