The app, screen by screen
Sandbox
Paste any text in and see precisely what Locke would do to it — what it finds, what it swaps in, and how sure it is. Nothing here is sent anywhere.
Available toAll plans
The Sandbox is the only place in Locke where you get to see the machinery work on demand rather than in passing. It's the right tool for four jobs: checking a new custom rule actually matches, finding out why something you expected to be caught wasn't, seeing what a stand-in looks like before you rely on it, and satisfying yourself that any of this works at all.
A run that found four things — including one caught by a custom rule.
- Input. Type or paste anything. It arrives pre-filled with a sample that trips several detectors at once.
- Masked output. Exactly what an AI tool would receive. Green highlights are the stand-ins.
- What was detected. One row per find: what it was, what replaced it, and the confidence score.
- Test custom rules decides whether your own rules run in this test. Leave it on to check a rule you just wrote.
Running a test
- Type or paste text into Input. Use real examples if you like — nothing here leaves your computer, and nothing is stored.
- Leave Test custom rules on to include your own rules, or switch it off to see only what the built-in detectors find.
- Press Run test.
- Read the results: highlights in both boxes, and a row per detection in the table below.
Copy masked output puts the masked version on your clipboard, which is handy when you want to paste something into an AI tool that Locke doesn't cover. + Add detection as rule jumps to Controls to turn what you just found into a permanent rule. Clear empties the box.
Testing a custom rule before you trust it
1You've written a rule for your firm's matter codes and want to know it works. Open the Sandbox and paste in a line that contains one, alongside some ordinary sensitive data.
Read all 4 steps as text
- You've written a rule for your firm's matter codes and want to know it works. Open the Sandbox and paste in a line that contains one, alongside some ordinary sensitive data.
- Press Run test.
- Locke highlights every value it found in amber, and lists them below — the name, the SSN, the email, and MTR-00421 caught by your custom rule. That last row is the confirmation you were after.
- The output box fills with the cleaned version: same sentence, same shape, different data. That is exactly what an AI tool would have received. Confidence for the custom rule is 100% — pattern rules either match or they don't.
Reading the results
| Column | What it tells you |
|---|---|
| Action | What Locke did. masked means replaced; flag rules are recorded without changing the output. |
| Type | Which detector matched — a built-in category, or one of your custom rules. |
| Replacement | The stand-in that went out in place of the original. |
| Confidence | How certain Locke is, as a percentage. Custom pattern rules read 100% because a pattern either matches or it doesn't. |
Highlight colours carry the same information visually: amber in the input for what was found, green in the output for what replaced it.
When it finds nothing
No detections in this input means Locke genuinely recognized nothing sensitive. Three usual reasons, in order of likelihood:
| Reason | Fix |
|---|---|
| The relevant data type is switched off | Check Controls › Detection data types. Region-specific types are off by default. |
| The value is below the confidence floor | Try Strict in Controls › Scanning floor and run it again. |
| It's an identifier only your organization would recognize | That's what custom rules are for — see Controls. |
