Problem
Detection over-catches by design (recall > precision; the safety net is over-detection + reveal-before-copy). But the user is the final authority on what's actually theirs to hide. Today the redaction set is all-or-nothing per run — no way to say "actually, keep this one."
Proposal
In the result view, let the user opt an individual detection OUT before Copy/Save:
- Click a
[REDACTED] span (or use its chip) to restore that occurrence's original value.
- The output, the counts ("N redacted · M types"), and Copy/Save reflect the kept set.
- Builds on the existing Reveal-originals (which already shows the truth in place); this makes a per-item choice persist into the saved document.
Notes
- Per-run keep-set only (originals are never stored; nothing persists across files).
- Distinct from Reveal-originals: a "kept" item is intentionally not redacted, vs. revealed-for-checking.
- A "keep all of this type" bulk action is the natural extension (see the companion request).
Source: user feedback during a result-view review.
Problem
Detection over-catches by design (recall > precision; the safety net is over-detection + reveal-before-copy). But the user is the final authority on what's actually theirs to hide. Today the redaction set is all-or-nothing per run — no way to say "actually, keep this one."
Proposal
In the result view, let the user opt an individual detection OUT before Copy/Save:
[REDACTED]span (or use its chip) to restore that occurrence's original value.Notes
Source: user feedback during a result-view review.