TIME investigation revealed that OpenAI paid Sama, a San Francisco-based outsourcing firm, approximately $200,000 (later stated as ~$150,000) to have Kenyan workers label tens of thousands of text snippets containing graphic descriptions of violence, sexual abuse, and hate speech to train safety filters for ChatGPT. Workers earning $1.32–$2 per hour reported psychological trauma from repeated exposure to extreme content; mental health support was inadequate. Sama terminated all OpenAI contracts in February 2022, eight months early, after TIME published a related investigation into Sama's work for Facebook, and subsequently exited all content moderation work entirely in January 2023. The episode surfaces structural dependency on low-cost human labor in the Global South to build frontier AI safety infrastructure, with cost passed to vendors and workers bearing uncompensated psychological risk.
219 other tracked developments touching the same claims.