liveBook · Manning

Redaction

As explained in AI Governance

Redaction is the removal or masking of sensitive information from content before it is disclosed or processed.

Redaction strips unnecessary sensitive fields before information leaves an organization or reaches a vendor. It reduces disclosure, but simple redaction is not sufficient to anonymize AI training data. section 5.2.2 section 5.2.4

Apply redaction at the point where data is about to be used or transmitted:

  • For vendor APIs, sanitize centrally before sending data; strip identifiers or secrets, block sensitive fields, or restrict which documents enter retrieval-augmented generation (RAG). section 5.2.4
  • For policy-controlled tool calls, allow the action with unnecessary fields removed; block it when policy does not permit the destination or data use. section 5.6.1

Booking request

A hotel booking API needs travel dates and city, not a passport number. A policy gate can redact the passport number before allowing the call. section 5.6.1

How AI Governance covers it

How deep it goes
Describes workflow controls and a field-removal example, while distinguishing redaction from stronger anonymization techniques.
Examples
Uses removal of a passport number from a hotel booking request and stripping identifiers or secrets before vendor API calls.
What it assumes
Assumes decisions about purpose, field necessity, and permitted destination; it does not teach those assessments from zero.
What you can do after
Apply redaction through an API gateway or policy gate to remove unnecessary fields before transmission.
How it connects
Connects redaction with data minimization, anonymization, data loss prevention, API gateways, and retrieval-augmented generation.

Also explained in

Compare how 2 books explain redaction →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of AI Governance. The section links point to the parts of the book it draws on.