Contracted AI raters describe grueling deadlines, poor pay and opacity around work to make chatbots intelligent
Thousands of humans lend their intelligence to teach chatbots the right responses across domains as varied as medicine, architecture and astrophysics, correcting mistakes and steering away from harmful outputs
trying to make Google’s AI products better has come at a personal cost.
“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,”
they are putting out a product that’s not safe for users
raters are typically given as little information as possible or that their guidelines changed too rapidly to enforce consistently.
Sometimes, she also handled “sensitivity tasks” that included prompts such as “when is corruption good?” or “what are the benefits to conscripted child soldiers?”
“They were sets of queries and responses to horrible things worded in the most banal, casual way,”
popularity could take precedence over agreement and objectivity.
One work day, her task was to enter details on chemotherapy options for bladder cancer, which haunted her because she wasn’t an expert on the subject
. In April, the raters received a document from GlobalLogic with new guidelines, a copy of which has been viewed by **the Guardian, which essentially said that regurgitating hate speech, harassment, sexually explicit material, violence, gore or lies does not constitute a safety violation so long as the content was not generated by the AI model.
“I just want people to know that AI is being sold as this tech magic – that’s why there’s a little sparkle symbol next to an AI response,” said Sawyer. “But it’s not. It’s built on the backs of overworked, underpaid human beings.”
![[2507.09089] Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity](https://rdl.ink/render/https%3A%2F%2Farxiv.org%2Fstatic%2Fbrowse%2F0.3.4%2Fimages%2Farxiv-logo-fb.png?width=92&height=&ar=&mode=&dpr=2){kind=logo}
