A new investigation by The Verge and New York Magazine suggests that artificial intelligence is not only displacing workers but increasingly hiring them back as contractors to train the systems replacing them.
The report describes a growing ecosystem of gig work where laid-off professionals—including lawyers, writers, teachers, and scientists—are paid to produce training data that helps AI systems perform the very tasks those workers once did.
In many cases, these workers are writing prompts, grading chatbot responses, and creating detailed “rubrics” that define what a correct answer should look like. That data is then used by AI labs to improve their models.
As one worker described the situation in the report: “My job is gone because of ChatGPT, and I was being invited to train the model to do the worst version of it imaginable.”
The worker, identified in the report as Katya, had previously worked in content marketing before automation reduced demand for her role. She later joined a platform that recruits professionals to help train AI systems. Companies including Mercor, Scale AI, and Surge AI are hiring thousands of professionals to create training datasets used by large AI models.
These workers perform tasks such as:
- Writing example chatbot prompts
- Evaluating AI responses
- Producing “ideal answers” known as gold outputs
- Designing tests meant to make AI models fail
The goal is to help companies like OpenAI and Anthropic improve their models.
Mercor alone says around 30,000 professionals work on its platform each week, according to the report.
David Adubiina
The workers recruited for these roles often have advanced degrees and specialised expertise. Job listings reviewed in the investigation sought professionals, including:
- lawyers
- management consultants
- teachers
- scientists
- archivists
- journalists
Some projects even requested experts in niche topics such as “North American early-teen humour" to help train AI systems to better understand slang and social trends.
Industry insiders describe the effort as one of the largest attempts to collect human expertise ever undertaken. Despite the technical nature of the work, many workers report unstable conditions. Projects can disappear suddenly, hours fluctuate unpredictably, and pay often declines over time as companies attempt to reduce costs.
One worker in the report described the situation bluntly: “I’m being handed a shovel and told to dig my own grave.”
Another compared the work environment to a gig economy similar to ride-hailing platforms, where tasks appear sporadically, and workers compete to complete them before others.
The system reflects a paradox at the center of the AI boom. While companies promote the idea of fully autonomous AI systems, they are currently spending billions of dollars on human labor to teach those systems how to perform complex tasks.
Yet the long-term goal of the work remains automation. For many workers, that creates a strange reality where their current job is to help build the technology that may eventually eliminate their profession. As one screenwriter interviewed in the investigation put it, "It's like being asked to dig your own grave.”
David Adubiina
