One leaked SSN. That's a breach notice, not a maybe.

Honest about redaction on scanned pages. On a text-layer PDF, redaction genuinely removes the underlying content — we verify this by re-extracting the output and confirming the data is gone, not just painted over. On a scanned or image-only page, our AI still finds the PII, but we don't yet have a reliable way to redact pixels we can't map back to a location on the page. Rather than hand back a document that looks redacted and isn't, the system refuses the operation and tells you exactly which findings need a manual pass. That's slower for scanned content today — and it's the honest answer, not an overlay pretending to be one.