Back to Gap Map Product / UX / Safety & Guardrails

Presidio

Data Privacy Stack
open source / Overall score: 3.5

Presidio finds and removes personal information such as names, phone numbers, credit card numbers and locations in text, images and structured data. It combines named-entity recognition, regular expressions, checksums and context rules with optional outside models, and can redact, mask, replace or encrypt what it finds. Microsoft created it, and the community organization Data Privacy Stack now maintains it.

Openness

5 high confidence
5.0
license
MIT (OSI)
source
public (the analyzer, anonymizer, image redactor and structured-data packages all build from the repository)
core features withheld
no — a community-governed project with no commercial owner and no enterprise path in the repository tree

Presidio's code is MIT-licensed, and every package from the text analyzer to the image redactor builds from the public repository. Since leaving Microsoft it has had no commercial owner, so there is no paid edition for features to be held back for.

Adoption

4 high confidence
4.0

PyPI downloads of presidio-analyzer, the package that finds personal information and the one every Presidio setup installs. The anonymizer and image packages are published separately and are not added, since they are mostly installed alongside the analyzer.

Capability

3 medium confidence
3.0

Presidio finds and strips personal information across text, images and tables, with far more control over what counts as personal data than a general safety classifier gives. Privacy is the only risk it handles, while Llama Guard screens prompts and replies across fourteen hazard categories, so it sits a step below it.

  • https://presidio.dataprivacystack.org/ recorded 2026-11-12

    The documentation home describes identification and anonymization of private entities in text and images, and links guides for transformers, GLiNER, Flair and Azure AI Language recognizers, encryption, pseudonymization, structured data and DICOM redaction.

  • https://raw.githubusercontent.com/data-privacy-stack/presidio/main/README.MD recorded 2026-11-12

    The README lists "Predefined or custom PII recognizers leveraging Named Entity Recognition, regular expressions, rule based logic and checksum with relevant context in multiple languages", "Options for connecting to external PII detection models" and a "Module for redacting PII text in images (standard image types and DICOM medical images)".

Verified 2026-11-12