OpenAI Is Paying Thousands of Freelancers to Train ChatGPT to Do Their Jobs
A new OpenAI initiative, “Project Stagecraft,” is paying between 3,000 and 4,000 freelancers at least $50 per hour — and up to $500 for domain experts — to build occupation-specific training data for ChatGPT across more than 400 distinct job titles (Source: Business Insider). The project runs through data-labeling startup Handshake AI, with contractors spanning fields from commercial aviation and emergency medicine to plant science, pharmacology, and agricultural management (Source: Startup Fortune).
Contractors are directed to develop detailed personas and simulate real workflows, providing "context, goals, references, and deliverables" to help train the models with specialized human expertise. A Handshake training guide reviewed by reporters indicates the focus is on "knowledge work, not manual labor," with the data collected intended to "map economically relevant tasks and evaluate the model's capabilities."
One contractor who participated told reporters, "We all were aware that we were basically training AI to replace us" (Source: Business Insider).
The revelation arrives as OpenAI has simultaneously begun drafting policy papers on economic disruption and "rethinking the social contract" in an age of AI-driven labor displacement (Source: Vanity Fair).
The company recently closed a record $122 billion funding round at an $852 billion valuation, with enterprise revenue now accounting for over 40% of its $2 billion-per-month run rate (Source: OpenAI).
Stagecraft suggests that OpenAI's ambitions extend well beyond code generation into a systematic cataloging of professional knowledge — field by field, task by task.

