Roles and groups
Existing organisations can be mapped through AD/LDAP or managed in the hub.
AI receives only what has actually been approved
AI permissions and data sovereignty down to source level.
PaulinAI checks users, roles and groups before content is searched or passed to a model. Blocked data never enters the answer context.
For SMEs & industry
Secure enterprise AI begins before the answer.
Redacting information afterwards is too late. PaulinAI places the security boundary before search and AI processing: unauthorised folders, databases, mailboxes and websites are excluded from the request.
Source-level control
Approvals apply specifically to folders, databases, mailboxes, websites and other sources.
Filter before search
Unauthorised content is neither retrieved nor passed to the AI.
Agent permissions
Specialised agents receive only the tools and data required for their task.
Controlled external services
Cloud models and web search can be disabled by data class and process.
Traceable changes
Permissions and responsible roles can be maintained under control.
Introduce with control
From a defined process to productive local AI.
- 01
Connect identities
Users, groups and roles are imported from existing structures or managed in the hub.
- 02
Assign sources
Every source receives an owner, protection class and permitted groups.
- 03
Test access
Permitted and blocked questions are tested systematically with different roles.
- 04
Review regularly
Role changes, new sources and changed processes are reflected in permissions.
Related solutions
Continue with the topics that fit your project.
Frequently asked questions
Clear answers.
Can a manipulated prompt reveal blocked sources?+
Unauthorised sources never enter the answer context, so rephrasing a question cannot retrieve data the user did not technically receive.
Can Active Directory be used?+
Yes. Users and groups can be connected through existing directory services without storing the AD password in PaulinAI.
Does the permission model also apply to agents?+
Yes. Agents, P-Apps and users operate with defined source and tool approvals.
Set clear boundaries for AI data access.
We assess identities, sources, roles and agent permissions for an accountable pilot area.