CyBox AI Guardian

Security teams configure Browser Guardrails in CyBox AI Guardian. Each detection rule can use one of four enforcement actions:

Monitor — allow the submission and record the policy event Warn — require user acknowledgement before proceeding Mask — replace detected sensitive values and verify the sanitized content before submission Block — prevent the prompt or attachment from being submitted

Policy ranking is enforced consistently: Block overrides Mask, Mask overrides Warn, and Warn overrides Monitor.

Built-in detection covers private keys, API keys, passwords, JWTs, connection strings, payment cards, SSNs, national IDs, bulk email addresses, and additional sensitive-data patterns. Organizations can also define custom regular-expression rules for business-specific data.

Privacy by design

Prompt text, file bytes, extracted file content, and raw detected values remain inside the browser.

When a configured policy event occurs, CyBox receives security metadata only, such as the detection type, severity, non-reversible fingerprint, file basename, AI product and domain, timestamp, policy outcome, and relevant user decision.

Unsupported, malformed, encrypted, oversized, timed-out, or otherwise unavailable file scans remain blocked until the attachment is removed.

Enterprise deployment

CyBox AI Guardian is designed for managed security and IT deployments. It supports Chrome Enterprise force installation, managed configuration, organization collector tokens, and attributed Activity Feed events.

Requirements

An active CyBox AI Guardian organization and a valid organization deployment configuration.

Learn more at https://cybox.ai