liveBook · Manning

Output integrity

As explained in AI Governance

Output integrity is the privacy-risk dimension concerning harmful or false statements about real people produced by AI systems.

Output integrity is the privacy risk that an AI system produces harmful or false statements about real people. A fabricated personal fact can cause harm when professionals rely on it without verification. section 5.0 section 5.1

Human review must involve proper scrutiny: approving AI output without checking it is not meaningful human oversight. Automation bias can lead people to rely on hallucinated claims and make unfair decisions. section 5.1 section 5.1

False or harmful claims about people can include:

How AI Governance covers it

How deep it goes
Defines the risk and explains how hallucinations, automation bias, and inadequate human review can contribute to harm.
Examples
Names invented medical histories, fabricated criminal records, and misattributed statements.
What you can do after
Evaluate claims about people with meaningful human scrutiny rather than approving AI output without proper verification.
How it connects
Connects privacy risk to hallucinations, automation bias, and human review.

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.