Edison Watch
The Data Firewall for AI Agents. Prevent exfiltration, gain visibility, and enforce security policies.
Edison Watch is a security gateway for AI agents. It monitors every action, enforces granular policies, and blocks data exfiltration in real-time.
The Problem: AI agents are accessing your enterprise data - calendars, documents, CRMs, databases - with no guardrails. One malicious prompt can exfiltrate sensitive information before you even know it happened.
How it Works: The Lethal Trifecta
Edison Watch prevents leaks by detecting the "Lethal Trifecta"—the combination of capabilities required for exfiltration:
| Capability | Risk |
|---|---|
| Private Data Access | Agent reads sensitive internal info. |
| Untrusted Content | Agent processes external content (web pages, etc). |
| External Communication | Agent sends data outside (Email, Slack, APIs). |
When all three are present, Edison Watch pauses the dangerous action for human review.
Get Started
Security Model
Understand how Edison's policy engine works - the three flags, ACLs, and enforcement logic.
Dashboard
Monitor activity, review sessions, and manage approvals.
Admin Guide
Manage users, roles, and server configurations.
IT & Security
Deploy Edison and understand the security architecture.
Core Features
- Trifecta Detection: Block exfiltration before it happens.
- Human-in-the-Loop: Approve risky actions in real-time.
- Audit Trail: Complete record of every AI tool invocation.
- Enterprise Ready: SIEM integration and SSO support.