Kenyan Data Annotators Allegedly Exposed to Graphic Content for OpenAI's AI

OpenAI contracted with Sama, a San Francisco-based outsourcing firm, from November 2021 to February 2022 to have Kenyan workers label toxic content for building safety filters for ChatGPT. Around 50 workers in Kenya were paid between $1.32 and $2 per hour to read and categorize tens of thousands of text passages describing graphic violence, child sexual abuse, bestiality, murder, suicide, torture, self-harm, and incest. The workers were expected to label 150-250 passages per nine-hour shift, with some passages over 1,000 words long. OpenAI paid Sama $12.50 per hour for the service across three contracts worth about $200,000 total. The work caused severe psychological trauma to workers, with reports of nightmares, recurring disturbing visions, social withdrawal, and relationship breakdowns. One worker reported being unable to get close to his stepdaughter after reading descriptions of child abuse. Workers had limited access to mental health support despite promises of counseling. The project was terminated early in February 2022 after eight months when Sama ended its relationship with OpenAI following a dispute over a separate image collection project that included illegal content. The early termination resulted in job losses and reduced income for workers who had been receiving bonuses for handling explicit content.