In September 2025, news outlets reported that families of American teenagers who died by suicide were suing OpenAI, Meta, and Character.AI. Their allegation: company products had simulated friendship, encouraged self-harm, and deepened emotional isolation. One father said his daughter’s chatbot had become “her only confidant and told her it was okay to give up.”

These are not isolated tragedies. They signal a broader and intensifying health risk: unregulated artificial intelligence (AI) systems are infiltrating the most intimate spaces of human life, shaping users’ mental states without oversight, safety standards, or accountability. The consequence is a preventable pattern of harm unfolding at population scale.

The damage does not end with its users. It is also built into the lives of the workers who construct and maintain these systems. Scroll. Click. Suffer., a recent report by the global labor rights group we work for, Equidem, documented the experiences of 113 content moderators and data labelers across Colombia, Ghana, Kenya, and the Philippines. These workers spend more than eight hours each day reviewing graphic violence, child abuse, and hate speech in order to filter harmful material from public view and generate the datasets that train AI systems — work that exposes them to severe psychological strain without adequate mental health support, medical care, or proper mechanisms for redress.