Underredaction

Underredaction is the insufficient removal or masking of sensitive information in a redacted document, resulting in the exposure of confidential or restricted data. It is the opposite of over-redaction, where too much information is removed. The term is used in privacy, security, and compliance contexts.

It commonly arises in government documents, legal filings, corporate communications, and research data releases where personal data, trade secrets, or national security information must be protected. Redaction aims to prevent unintended disclosure while preserving the usefulness of the remaining content.

Causes of underredaction include imperfect data classification, ambiguous or overlapping data elements, and heavy reliance on automated redaction tools that lack comprehensive recall or rules. Human error, fatigue, time pressure, and the presence of sensitive information in nontext fields or embedded metadata can also contribute.

Examples of underredaction include a legal or FOIA release that still contains unmasked personal identifiers such as names or account numbers, a contract with client details visible, or documents where metadata or OCR artifacts reveal sensitive information that was not properly redacted.

Consequences can be significant, including privacy violations, regulatory penalties, potential lawsuits, harm to individuals, and reputational damage to the releasing organization. In the research and public sector, underredaction may undermine trust and impede compliance with data protection laws.

Mitigation strategies emphasize robust redaction policies and processes. These include using high-recall automated redaction tools in combination with human review, conducting redaction audits, minimizing data collection, removing metadata, and maintaining an audit trail to verify that redactions are complete and correct. See also redaction, data privacy, de-identification.