Starting point
We run our own AI. Other AI services should be blocked.
Matching policy
Allow internal AI and block other detected AI connections.
Start with: Protect
Patronus Monitor & Protect
AI security starts on the company computer.
Patronus runs on the device where you use AI. Monitor discovers that activity. Protect also checks content and enforces your rules.
What Protect does
Four things Patronus handles on the device.
Real views from the product. Everything runs locally, none of it needs an additional cloud service.
System-wide coverage
One agent for every AI tool in the team.
Patronus works at the application layer, not inside a browser extension or a single app. Every AI request the device makes is captured, regardless of tool, provider or application. Including AI nobody approved.
Local architecture
Your data stays on the device.
The whole check runs on the machine. No prompt is sent to another vendor for analysis. It works offline and in isolated networks, and GDPR is met by the architecture, not by configuration.
Real-time protection
Caught before it reaches the model.
Prompt injection, jailbreak attempts, personal data and policy violations are detected and stopped on the device within milliseconds. The workflow stays smooth.
- Prompt injection filtering
- Redaction of personal and sensitive data
- Jailbreak and manipulation detection
- Data exfiltration prevention
- Content policy enforcement
- Detection of unapproved AI
- Control of agent actions
Policy enforcement
You set the rules. Patronus enforces them everywhere.
Define company-wide rules that apply to every tool: block data types, restrict behavior, limit what agents can access. Every decision is logged locally.
Example policies
What do you want to make possible with AI?
Start with your goal. Each card shows a typical starting point and the policy that fits.
Starting point
We approve selected AI services. The rest should stay blocked.
Matching policy
Allow approved destinations. Block other detected AI destinations.
Start with: Protect
Starting point
Our team uses internal AI and selected external services.
Matching policy
Allow both paths and add rules for sensitive content.
Start with: Protect
Starting point
We want to keep Chrome or another app without approving every new AI feature.
Matching policy
Check together whether the specific AI feature can be detected and controlled separately.
Start with: Protect · Fit-Check
An everyday example
One prompt. Three points of control.
An employee asks AI to draft a customer email. The draft includes confidential contract details.
Illustrative example · fictional data
See which AI tool is being used and which application the request comes from.
AI activity discovered
- Application
- Browser
- Activity
- Request to ChatGPT
- Status
- Discovered with Monitor
Example input (not a Monitor finding)
Draft an email about the Northlight contract. Our internal margin is 24%.
What do you need to control?
Sensitive content
Set rules for customer data, internal information and other content that needs protection.
Protect your data →Traceable decisions
Use inventory and logs to support your documentation.
Explore records & evidence →FAQ