What is Sentinel?

Overview

Sentinel Agent is a lightweight system service / headless daemon that automatically monitors your employees’ AI tool usage. It logs which AI services they access (ChatGPT, Claude, Copilot, etc.), what they’re doing, and automatically enforces policies.

Sentinel fleet management dashboard for endpoint governance

Use Cases:

Prevent employees from pasting confidential data into ChatGPT
Audit which employees use AI and how
Enforce approved AI tools only
Prevent data loss and IP theft

How Sentinel Works

Install: Deploy to Windows, Mac, or Linux
Monitor: Observes all AI tool access (ChatGPT, Claude, etc.)
Enforce: Blocks unapproved tools or sensitive data
Report: Central dashboard shows usage

Employee → ChatGPT
    ↓
Sentinel Agent (local)
    ↓
[Policy Check: Is ChatGPT approved?]
    ↓
If approved: Allow + Log
If blocked: Block + Alert admin
    ↓
TruthVouch Dashboard (admin view)
    ↓
Reports & alerts

Sentinel vs. Firewall

Feature	Sentinel	Firewall
Scope	Desktop employee usage	Application/API usage
Tools	ChatGPT, Claude, Copilot web	Backend AI models
Deployment	Desktop app	Proxy/sidecar
Data	Screenshots, URLs, copy/paste	Requests/responses
Use Case	Employee governance	Application governance

Key Features

Usage Monitoring

Track which employees use AI and when:

Dashboard shows:
- Users: 150 active last month
- Tools: ChatGPT (80%), Claude (15%), Copilot (5%)
- Time: Average 2.3 hours/week per user
- Trending: Usage up 15% month-over-month

Policy Enforcement

Block employees from using unapproved tools or pasting sensitive data.

Policy: "Prevent PII to ChatGPT"
  Action: If employee tries to paste SSN → Block
  Alert: Admin notified
  Result: Data prevented from leaving company

Approved Tool Whitelist

Only allow specific AI tools:

Approved:
  ✓ ChatGPT-4 (internal account)
  ✓ Claude (enterprise)
  ✓ GitHub Copilot (licensed)

Blocked:
  ✗ ChatGPT free tier
  ✗ Any other LLM

Data Loss Prevention (DLP)

Block sensitive data from being sent to public AI:

Sensitive patterns detected:
  - Credit cards
  - Passwords
  - Source code
  - Customer PII

Action: Block input to public LLMs

Compliance Reporting

Pre-built reports for audits:

Reports:
- Who used AI and when
- What data was blocked
- Policy violations
- Certified safe usage

Deployment Options

SaaS Managed

TruthVouch hosts everything:

Easy installer (MSI, DMG, deb)
Auto-updates
Cloud-based management
No infrastructure required

Best for: Most organizations, quick deployment

Self-Hosted

Deploy on your infrastructure:

Full control
Air-gapped possible
Custom integrations
Manage updates yourself

Best for: Highly regulated, strict data residency

Installation

Choose your platform:

Windows — MSI installer, system service
macOS — DMG or Homebrew
Linux — deb/rpm packages, systemd

Installation takes <5 minutes.

Policy Configuration

Set up policies before rollout:

Go to Governance → Sentinel → Policies
Approved Tools: List what’s allowed
DLP Rules: What data to block
Scope: Which users/departments
Actions: Block vs. warn
Deploy to devices

Example Policy:

Name: "Standard Employee Policy"
Approved Tools:
  ✓ ChatGPT-4 (company account)
  ✓ Claude (via TruthVouch)

Blocked:
  ✗ ChatGPT free tier
  ✗ Any unauthorized LLM

DLP Rules:
  - Block: SSN, passwords, API keys, credit cards
  - Allow: General questions, public information

Scope: All employees except IT department
Action: Block (not just warn)

Monitoring & Reporting

Central dashboard to see all usage:

Go to Governance → Sentinel → Dashboard
See metrics:
- Active users
- Tool usage breakdown
- Policy violations
- DLP blocks
- Top users/tools
Generate reports for compliance:
- Weekly usage summary
- Monthly compliance report
- Incident investigation

Employee Experience

Employees see transparent enforcement:

Normal Use:

Employee opens approved tool
Sentinel allows
Work continues

Blocked Attempt:

Employee tries to use unapproved tool
Sentinel blocks with message: “This tool is not approved by your organization”
Option to request access (if configured)

DLP Block:

Employee tries to paste credit card into ChatGPT
Sentinel blocks: “Sensitive data detected. This action is not allowed.”
Data is not sent anywhere

Integration

Okta/AD Integration

Automatically get user lists and groups:

Okta Group: Designers
  → Assigned policy: "Design Team Policy"
  → Approved tools: ChatGPT-4, Claude
  → DLP: Block code, block passwords

Slack Notifications

Get alerts on important events:

Channel: #security-alerts
Message: "[email protected] attempted to paste company password into ChatGPT - BLOCKED"

SIEM Integration

Send logs to security systems:

Splunk/ELK receives:
- All policy violations
- Tool usage
- DLP blocks
- Searchable & alertable