Detect
DMZ looks for selected identifiers and secrets in supported ChatGPT and Claude text submissions. Processing happens on your device.
// A FREE COMMUNITY PROJECT BY KASTRO
Fewer identifiers. More care..
DMZ attempts to mask selected identifiers before you send text to ChatGPT or Claude. It processes locally on your device. A free tool focused on Icelandic text, with clear limitations.
The DMZ Chromium extension is available for testing with fabricated data. ZIP package, not a browser-store release.
extension · synthetic data
01 · You typeScanning…
Arna Jónsdóttir, kt. 140182-8858, has requested a quote. Draft a reply I can send to demo@example.com.
02 · Model provider receivesIdentifiers replaced
⟦NAME-1: XXXXX XXXXX⟧, kt. ⟦KT-2: XXXXX-XXXX⟧, has requested a quote. Draft a reply I can send to ⟦EMAIL-3: XXXXX@XXXXX.XXX⟧.
03 · You readModel reply · redacted
Hello , thank you for your enquiry. We will send you the quote tomorrow.
Session map · on this device · cleared with sessionArna Jónsdóttir⇄⟦NAME-1: XXXXX XXXXX⟧140182-8858⇄⟦KT-2: XXXXX-XXXX⟧demo@example.com⇄⟦EMAIL-3: XXXXX@XXXXX.XXX⟧
Illustration of planned behaviour; actual markers also include a random session ID. All values are synthetic and test kennitölur use century digit 8. The working DMZ prototype uses a manual side panel.
Detect. Mask. Restore.
An aid within your chat, with a local session map.
Mask
Detected values become visible markers. Always review the text: some identifiers can escape detection.
Restore
When unchanged markers return, DMZ restores matching values in a local reply view. The mapping stays on your device.
Useful without a sales conversation.
DMZ follows the free community direction of Dmark and Kastro Lookup.
Free to use
No DMZ subscription, seat minimum or paid security tier. No DMZ account needed. Your AI service’s costs and terms are separate.
On your device
DMZ sends neither your prompt nor its mapping to DMZ. The AI site receives the text you submit, including anything missed.
Built with feedback
Feedback, fabricated test cases and translations help improve the tool. Kastro assistance is optional.
UNDERSTAND THE LIMITS
Review before you send.
DMZ can miss identifiers. On failure, submission continues with a warning. Files, images and attachments are sent unchanged with a warning. Context can still identify people; masking does not guarantee anonymity.
Read the data boundaries →DMZ · new test build · 05.10.2026
More identifiers masked in full.
A new private test build improves masking of email addresses, credentials and supported numbered street addresses without postcodes. Restoration preserves the original text, including accented characters, and pending scans are handled more reliably. A draft edited while waiting is not sent automatically.
These changes are in a test build, not a general release. The name model is unchanged and the 99% detection target remains unproven.
Name detection · development status · 05.10.2026
The name model remains unchanged.
The latest training round did not produce a model that passed comparison with the current model. Better recall on synthetic legal and financial text came with lower recall on general text. The existing model remains in the separate DMZ test build. There is no general release.
| Metric | General text · 5,582 excerpts | Synthetic legal/financial text · 256 documents |
|---|---|---|
| Excerpts with every name correct | 98.57% | 97.66% |
| Name recall | 97.01% | 98.75% |
| Detection precision | 95.55% | 99.37% |
| Missed name spans | 45 | 6 |
| False detections | 68 | 3 |
The target is at least 99% exact entity-set accuracy and 99% recall in each primary text category. It has not been reached. Independent validation is still pending.
An excerpt is correct only when every annotated entity has the right type and exact span, with no extra detections. Recall measures the share of annotated entities found. Precision is a different metric: the share of detections that are correct. This table evaluates names only; the acceptance target covers all entity types.
These are development results after repeated model and parameter selection, not independent final validation. The synthetic documents reuse templates and annotation review is pending. These figures do not establish everyday performance or the accuracy of DMZ as a whole.
FABRICATED EXAMPLES · USEFUL FEEDBACK