India's female workers watching hours of abusive content to train AI

3 months ago

Monsumi Murmu, a 26-year-old content moderator in India, works from her village, classifying violent and disturbing content flagged by AI systems.
Content moderation involves viewing up to 800 videos/images daily, leading to emotional numbing and psychological trauma.
Studies show content moderators experience traumatic stress, intrusive thoughts, and sleep disturbances, with lasting cognitive impacts.
India's data annotation sector employs ~70,000 workers, mostly from rural/marginalized backgrounds, with 80% being women.
Women are preferred for their perceived reliability and willingness to accept home-based work, but this reinforces marginalization.
Workers often face unexpected exposure to graphic content (e.g., child abuse, pornography), causing personal distress and dissociation.
Job listings are vague, and NDAs prevent workers from discussing trauma, isolating them further.
Psychological support is rare, and India’s labor laws lack protections for mental health harms in this sector.
Workers like Murmu cope through solitary activities (e.g., walks, painting) but fear unemployment more than the job’s toll.

Hasty Briefsbeta