PII Masking Health Data Preview

AI4Privacy's PII-Masking-2M European release provides a preview of 50 sample entries from a dataset for masking Personal Health Information. The full dataset contains 200,000 entries, focusing on European coverage. This preview was last updated in April 2026.

Use Cases

• Benchmarking PII detection models using the labeled source text and redacted PII values.
• Analyzing label distribution across different PII categories present in the health text.
• Studying language distribution patterns within European health data.
• Training text masking algorithms on the structured redaction of sensitive health entities.

Strengths

• Preview contains 50 sample entries from a larger 200,000-entry dataset.
• Dataset focuses on European health data, providing regional specificity.

Limitations

• Preview size is extremely small (50 rows) for meaningful model training.
• Source text and PII values are redacted in the preview, limiting direct analysis.
• Geographic coverage is limited to Europe, which may not generalize to other regions.

Provenance

Source

AI4Privacy.

Collection Method

Part of the PII-Masking-2M release; specific collection method unknown.

Freshness

Last updated April 2026.

Geography

European coverage.