Limitations
DataAnonymiser is a best-effort tool. Read this page before using it for high-stakes decisions.
PII detection is not exhaustive
Deterministic rules and the local processing engine will miss some sensitive entities, especially novel formats, uncommon names, or context-dependent identifiers. The residual-risk report is an aid, not a guarantee.
No pixel-level image redaction
For images, the app produces an anonymized textual description. It does not blur, redact, or modify the source image file. Do not share the original image.
No formal de-identification standard
This product does not claim compliance with HIPAA Safe Harbor, GDPR pseudonymization standards, or any other regulatory framework. Use it as a practical aid, not as a compliance tool.
Model quality is bounded
The local processing engine runs on CPU by default. Long documents may process slowly. Output quality depends on the installed processing pack version. Keep the app and local processing files updated for best results.
Retry is not perfect
If the validator detects residual leakage and retries with a stricter prompt, the retry may still not catch all sensitive details. Residual-risk warnings in the output pane mean the output requires your review.
Supported file types
Version 1 supports plain text, common source-code and developer config files, PDFs (.pdf), Word documents (.docx), spreadsheets (.xlsx), and images (.png, .jpg, .jpeg, .webp).