CyBox AI Guardian

Detects sensitive prompts and file uploads on allowlisted AI web tools. Enforces allow, warn, or block policy. Never sends content.

CyBox AI Guardian helps organizations govern how employees use web-based AI tools with local detection, real-time policy enforcement, and security visibility in CyBox AI Guardian. How it works The extension runs only on allowlisted AI websites (including ChatGPT, Claude, Gemini, Microsoft Copilot, Perplexity, and others). It scans prompts and supported file uploads locally in the browser when a user submits content or uploads a file, not on keystrokes, page loads, or general browsing. Organization policy enforcement Security teams configure Browser Guardrails in CyBox AI Guardian Policy. Each detection rule can be set to: Allow — log silently to the Activity Feed Warn — prompt the user to acknowledge before proceeding (when alerts are enabled) Block — stop prompt submission; record policy violations for file uploads Built-in detectors cover private keys, API keys, JWTs, connection strings, credit cards, SSNs, national IDs, bulk email addresses, and more. Organizations can add custom regex rules for domain-specific patterns. Privacy by design CyBox does not receive prompt text, file contents, or raw secret values. Nothing is sent unless a sensitive pattern is detected and the extension is configured with your organization's CyBox Guardian token and work email. When a detection occurs, CyBox receives metadata only — pattern type, severity, non-reversible fingerprint, file basename, AI product/domain, timestamp, policy outcome, and optional Allowed/Denied user decision. Enterprise deployment CyBox AI Guardian is designed for security and IT teams. Deploy via Chrome Enterprise force-install and managed storage policy. Users confirm their work email once for Activity Feed attribution. Requirements CyBox AI Guardian organization account https://cybox.ai