Trust-Prompt Enterprise
A structured technical overview of the Trust-Prompt Enterprise Admin Policy Builder, including policy targets, AI access control, scope settings, runtime protection, permissions, validation and export workflow.
Google Admin Policy Builder
Build managed Chrome policies for the Trust-Prompt Enterprise v0.2.3 baseline.
Core enterprise control layers
Trust-Prompt checks and controls AI usage before data leaves the device.
AI Access Control
- Define approved AI tools
- Block selected AI sites
- Apply controls per Policy Target
Prompt Protection
- Detect sensitive data locally
- Warn or block before send
- Support redaction behavior
Centralized Management
- Build managed policies
- Export Google Admin JSON
- Export Windows REG
Save or reopen builder projects locally
Projects are saved as local JSON files. No policy data is uploaded.
Policy Targets and group-based governance
Each target can represent a Google Organizational Unit, Google Group, Microsoft Entra group, department or custom team.
Reference profiles for faster setup
Preset cards provide starting points for common governance profiles.
Light
Softer Enterprise baseline for pilots.
Standard
Reference Enterprise baseline aligned with QA matrix.
Strict
High-control posture with stronger restrictions.
Finance
Finance-oriented profile for regulated workflows.
Support
Support team profile with selected controls.
IT Admin
Broad visibility and managed override permissions.
Runtime protection settings
These settings define the runtime baseline for the selected Policy Target.
Where Trust-Prompt can run checks
Scope controls where Trust-Prompt is active and how the browser extension reacts to each supported AI host.
Which AI tools each target may use
AI Access Control decides which AI tools are approved or blocked for the selected Policy Target.
Access modes
- Monitor approved sites only
- Block listed sites
- Allow approved sites only
Target-based controls
- Approved sites per target
- Blocked sites per target
- Custom AI blocklist
High-level risk management
Administrators assign ALLOW, WARN or BLOCK actions to enterprise risk categories.
FINANCIALBANK_ROUTING_IDENTIFIERSID_KYCLEGAL_COMPLIANCEPIISECURITYFine-grained controls and vocabulary tuning
Administrators can override specific detectors and tune organization-specific terminology.
Rule-level overrides
- IBAN detection
- API key and token detection
- Large paste and document markers
- ALLOW / WARN / BLOCK overrides
Word policies
- Allowlist reduces false positives
- Blocklist protects internal terms
- One term per line
User visibility and managed override permissions
Admins control what users can see and what users are allowed to change locally.
Visibility
- License section
- Governance controls
- Audit section
- Settings lock
- Scope section
Editing rights
- Protection toggle
- Upload policy
- Redaction
- Scope entries
- Rules and word policies
Policy health before deployment
The builder validates JSON readiness, REG readiness and policy conflict notes before export.
Deployment summary and export formats
Admins can export Google Admin JSON, Windows REG, and a human-readable Policy Summary.
Google Admin JSON
- Chrome managed storage policies
- Reflects selected Policy Target
- Includes runtime, scope, AI Access, categories and permissions
Windows Registry Export
- Windows-based Chrome or Edge deployments
- Requires valid Extension ID
- Preview appears when required fields are valid
Predictable policy priority
Trust-Prompt Enterprise applies policies in a consistent enforcement order.
1. Scope LOCK
Full interface restriction takes precedence.
2. Scope BLOCK
Access or send action is blocked.
3. AI Access BLOCK
Target-specific AI tool restriction is enforced.
4. Prompt Pre-check
Sensitive content is evaluated locally before send.
Local-first architecture
The product is designed to evaluate prompts locally without uploading prompt content.
Privacy principles
- Prompt content is evaluated locally
- No backend is required for enforcement
- No prompt content is stored by design
- Audit should remain metadata-only
Current limitations
- One selected Policy Target per export
- Custom hosts may require host permissions or Chrome Admin URL policies
- No OCR or file-content inspection unless separately implemented
