Find the sensitive databefore you share the file.

No — and this is the most important thing to understand about it. It is a readiness gate, not a redactor. It scans a document's text, points to exactly what's sensitive and what's still exposed, and gives you a CLEAR / REDACT FIRST / DO NOT SHARE verdict — but it never removes or masks anything itself. You do the redaction in your own tools, and you remain responsible for what you share. It tells you what to fix; it doesn't fix it for you.

It runs a pattern pack of ten deterministic detectors across three sensitivity tiers: high (US Social Security numbers, credit/debit card numbers, bank routing, health/medical markers), medium (labeled dates of birth, US street addresses, employer ID/TIN), and low (email addresses, phone numbers, IP addresses). Each finding is a fixed regular-expression match on the document's own text — no model, no guessing — so the same document always produces the same findings. The pattern pack and tiers are editable for your own identifiers and documents.

Because of the exposure gate. Any high-sensitivity finding left open — not yet marked redacted or allowed — forces DO NOT SHARE regardless of how high the coverage score is. In the worked example a customer record has its card and date of birth handled (75% coverage) but one SSN still open, and that single unhandled high finding blocks it. One open SSN is enough to expose a person, so it's dispositive. The coverage score is context only; the gate decides.

No. A CLEAR verdict means none of the patterns you've defined matched an unhandled finding — it is not proof the document is PII-free. A detector can miss data it wasn't written to catch (an unusual format, a foreign identifier, PII embedded in an image), so CLEAR is a strong signal, not a certification. It is not a certified or guaranteed-complete redaction tool, and it is not legal or compliance advice. Treat it as a deterministic safety net you tune, not a final sign-off.

When you scan a set of documents, the per-document verdicts roll up into one library verdict: SAFE TO RELEASE (every document is CLEAR), REDACT FIRST (some need redaction but none are blocked), or HOLD (at least one document is DO NOT SHARE). One blocked document holds the whole set — so a single file with an open SSN keeps the batch on HOLD until it's handled, rather than letting it ship alongside the clean ones.

They're opposite-facing halves of the document line. The Document Field Validator checks whether a record's fields are correct (formats, ranges, math) and rejects bad data. This scans assembled document text for sensitive content that must be handled before the file is shared or stored. One asks 'is this data right?'; the other asks 'is it safe to send?'. They pair: validate the fields by rule, then scan the same documents for PII before they leave the building. It also pairs with the AI Governance & Acceptable Use Starter Kit, which sets the policy this kit operationally checks.