Thousands of contractors are training Google's Gemini chatbot, and many describe poor working conditions and exposure to disturbing content, according to reporting by The Guardian. The workers, often called "AI raters," instruct the model and correct its mistakes — a role that The Guardian reports can involve moderating graphic material without prior warning.
Rachel Sawyer, contracted as a "generalist rater" for Google last year, told The Guardian she was "shocked" to be assigned "distressing content." She said she received no warning and was never asked to sign consent forms during onboarding, and that neither the job title nor description mentioned content moderation.
Raters without specialized expertise are also asked to verify accuracy in complex fields such as architecture, astrophysics, and medical guidance. One rater told the newspaper she was pressured to work faster while training the AI on sensitive medical topics like chemotherapy treatments for bladder cancer. "I pictured a person sitting in their car finding out that they have bladder cancer and googling what I'm editing," she said.
Japanese contractor GlobalLogic hired thousands of raters in the US to train Google's AI, according to The Guardian. Their hourly wages of $16 to $21 are far higher than counterparts in Africa and South Africa, the newspaper reported, but still below what some raters consider their work is worth. "They are people with expertise who are doing a lot of great writing work, who are being paid below what they're worth to make an AI model that, in my opinion, the world doesn't need," one rater told The Guardian.
Google disputed the idea that raters are responsible for making its AI models sound intelligent. A spokesperson told The Guardian that "quality raters" were "employed to provide external feedback on our products," which "help us measure how well our systems are working, but do not directly impact our algorithms or models."
Adio Dinika of the Distributed AI Research Institute described the arrangement to The Guardian as "a pyramid scheme of human labor." "These raters are the middle rung: invisible, essential and expendable," Dinika said.
Raters also face tight deadlines and rapidly changing guidelines. "We had no idea where it was going, how it was being used or to what end," a GlobalLogic-contracted rater told The Guardian. Despite their efforts, frontier AI models including Gemini continue to hallucinate, The Guardian noted.
Editorial Notes