Microsoft-Presidio: Mask Names, Credit Cards, and More in Your Data (github.com)

Microsoft’s Presidio is an open-source, extensible PII de-identification toolkit for both text and images that identifies and anonymizes sensitive data—names, credit cards (with checksum support), SSNs, locations, phone numbers, bitcoin wallets, financial data and more. It offers predefined and customizable recognizers that combine Named Entity Recognition, regular expressions, rule-based logic and contextual checks, supports multiple languages, and can plug in external detection models. Presidio also handles image redaction (standard formats and DICOM medical images) and can run across Python or PySpark workloads, in Docker containers or Kubernetes clusters, enabling fully automated or semi-automated de-identification flows for different operational needs. For the AI/ML community Presidio matters because it lowers the barrier to privacy-preserving data handling and governance—critical for regulatory compliance, responsible model training, and safe data sharing. Its pluggable architecture allows teams to tailor detection rules or integrate stronger models before training or release, reducing PII leakage and audit risk while preserving utility. The project includes demos, examples, docs and contribution workflows (CLA required) to encourage adoption and improvement, making it a practical building block for production data pipelines and privacy-first ML workflows.