Platform · Security
Intelligence without losing control.
Connecting AI with organizational information requires more than powerful models. It requires boundaries.
The path of every request
From identity to audit.
- 01IdentityWho is asking
- 02AuthenticationVerified sign-in
- 03PermissionsWhat they may see and do
- 04Data AccessOnly approved, permitted data
- 05ReasoningAnalysis within those bounds
- 06Action PermissionIs this action allowed?
- 07Human ApprovalWhere appropriate
- 08ExecutionControlled, scoped
- 09AuditRecorded and reviewable
Action permission, human approval and execution apply where Agent capabilities are enabled. Answering a question never changes your systems.
Principles we build to
AI should understand what it can do, and what it cannot.
WhoAccess
- Authentication and authorization
- Every request is tied to a verified identity.
- Role-based access
- What someone can ask about follows their role.
- Tenant isolation
- One organization's data and context stay separate from another's.
WhatData
- Secure data connectivity
- Connections limited to the data you choose to connect.
- Permission-aware retrieval
- Answers draw only on information the person is allowed to see.
- Encryption
- Data protected in transit and at rest.
- Responsible data handling
- Clear rules for what is stored, for how long, and why.
HowAction
- Controlled actions
- Agent works through scoped tools, not open-ended system access.
- Human approval where appropriate
- Actions that change things can require a person to approve them.
- Auditability
- Questions, tool calls and approvals can be reviewed afterwards.
- Deployment controls
- Choices about where QuerySafe runs, moving toward your own infrastructure.
Standards and credentials
Recognized practices behind how we build and deliver.
Independent standards and assessments support disciplined security, delivery and quality practices across the organization behind QuerySafe.
Certification and appraisal scope can apply to specific legal entities, teams, services or delivery processes. Current supporting documentation is available during evaluation.
FAQ
Security questions buyers ask
Does QuerySafe change our data when it answers a question?
Answering a question does not change your systems. QuerySafe Agent can perform approved actions, with permission checks and, where appropriate, human approval.
How are permissions enforced?
Every request is tied to a verified identity. What someone can ask about follows their role, and answers draw only on information that person is allowed to see.
Is one customer's data kept apart from another's?
Yes. One organization's data and context stay separate from another's, and QuerySafe connects only to the datasets you choose.
Can we review what was asked?
Yes. Questions, tool calls and approvals can be reviewed afterwards, so security and audit teams can see how QuerySafe was used.
Which certifications and assessments do you hold?
The credentials shown on this site are Cyber Essentials Certified, CMMI Maturity Level 3 and ISO 9001:2015. Scope can depend on the applicable legal entity, service and delivery process, so we provide current supporting documentation during evaluation.
Where does QuerySafe run?
Today QuerySafe runs as a managed service on Google Cloud, close to your BigQuery data. Private deployment inside your own cloud boundary can be discussed for organizations with stricter requirements.
Have security questions?
We'll walk your security and data teams through how QuerySafe connects, what it can see and what it stores.