S0022RS7-T23-S0022-Z · Full risk code

Imbalanced Labor Conditions in Data Annotation

資料標註勞動條件失衡

Inception
Risk Description

When model training and improvement depend on large volumes of low-wage data annotation and content moderation work; due to unmitigated control gaps, such workers are concentrated in lower-cost regions and face low wages, long hours, limited labor protection, and mental health impacts from prolonged exposure to harmful content, triggering external stakeholder impacts and causing these conditions are frequently obscured behind the technology's outputs.

Framework Mappings

NIST AI RMFGOVERN 1.1
ISO/IEC 42005影響評估流程
ISO/IEC 23894§6.3
ISO/IEC 5338啟動
MIT AI Risk RepositoryDomain 6

Risk Treatment & Implementation Guidance

Impose labor standards across the supply chain, gaining visibility into end-worker conditions under multi-tier outsourcing; Provide data-labeling workers mental-health support and reasonable-hours protection; Verify labor conditions via supply-chain due diligence and disclose publicly