S0084RS4-T16-S0084-Z · Full risk code

Unfair Group Representation and Erasure

群體表徵不公與消抹

Verification & Validation
Risk Description

When a generative system consistently depicts a single profile when portraying certain occupations or roles, rendering other groups nearly invisible; due to unmitigated control gaps, once such material is widely used, it reinforces rather than challenges existing societal frames of perception, triggering compliance exposure and operational reputational costs.

Framework Mappings

EU AI ActArt.10
NIST AI 600-1Harmful Bias and Homogenization
NIST AI RMFMEASURE 2.11
ISO/IEC TR 24027
ISO/IEC 5338驗證與確效
MIT AI Risk RepositoryDomain 1

Risk Treatment & Implementation Guidance

Audit representational fairness, quantifying group visibility and portrayal in generated content; Adjust data and generation strategy under inclusive-design principles to improve plurality; Gate widely circulated uses with representation-diversity checks