GlobiGuard Detect icon

GlobiGuard Detect

Detect and optionally redact sensitive data before it reaches models or external systems

GlobiGuard Detect

AI-generated

Summary

Scan each input item’s text with GlobiGuard using the selected industry profile, then route the item according to the detection result and optionally redact detected spans.

Inputs

  • Text (required) — Text to scan; defaults to the item’s text field. Empty or whitespace-only text is not sent for scanning.
  • Industry (required) — Domain profile for detection; defaults to GENERAL. The value is trimmed, uppercased, and must start with a letter followed by up to 63 letters, numbers, underscores, or hyphens.
  • Redaction Strategy — Choose Detect Only (none, the default), Drop detected spans, replace them with asterisks (mask), or replace them with [FIELD_TYPE] labels (replace).

Output shape

Three output branches: sensitive, clean, and error.

Successful scanned items are routed to sensitive when GlobiGuard reports detected fields or a non-ALLOW decision; otherwise they go to clean. Each retains its input data and gains detection details under json.globiguard, including the decision, entities, and redacted text when redaction is selected. Empty text goes to clean with scanned: false and reason empty_text. Errors are thrown unless Continue On Fail is enabled, in which case they go to the error branch with a sanitized error message. Text over 100,000 characters is rejected.

Links

Discussion