Data anonymization irreversibly modifies or removes personally identifiable information (PII) so individuals can no longer be identified. It can protect privacy when sensitive data is processed by generative AI, but it is not automatically sufficient: surrounding context may still reveal who was anonymized. section 8.6.2 section 1.3.6
Common anonymization techniques include:
- Data masking replaces sensitive values with fictional or synthetic values that preserve format and structure without connecting to the originals. section 8.6.2
- Aggregation combines individual records into groups so results show overall trends rather than individual-level information. section 8.6.2
- Generalization replaces precise values with broader categories or ranges, while noise addition alters values to obscure originals and preserve broad relationships. section 8.6.2
Anonymization versus pseudonymization
Anonymization removes or irreversibly changes identifying information. Pseudonymization instead replaces it with tokens, encryption, or other artificial identifiers that can be linked back to the original data; it is therefore reversible. section 8.6.2
