Search the Atlas

Search risks, controls, and glossary terms

MediumDefamationRealized

Defamation / reputational harm

Content Safety & Integrity

Description

The model asserts false, damaging statements about real people or entities.

Example scenario

An automated profile summary states a false adverse fact about a named borrower.

Real-world evidenceRealized

The Air Canada chatbot case (BC CRT 2024) and the FTC's DoNotPay settlement (2023) are both confirmed production instances where AI customer-facing systems caused consumer harm through deceptive or inaccurate representations, triggering legal liability or regulatory action under consumer protection frameworks.

Primary mitigations

  • Factual grounding
  • named-entity claim verification
  • defamation guardrails
  • human review of published content.

Detection signals

Unverified-claim detection on named entities; complaint monitoring.

Mitigating controls

4
Non-agentic controls

Related risks in Content Safety & Integrity