Imbalanced Labor Conditions in Data Annotation
資料標註勞動條件失衡
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
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