An AI startup powering Meta and OpenAI cut thousands of workers — then offered them a similar project for less money (www.businessinsider.com)

Mercor, a high‑growth data‑labeling startup used by major AI firms, abruptly ended a large Meta contract codenamed Musen, cutting thousands of contract labelers who had been doing work like ranking model prompts and tagging multimedia. Contractors said they were then offered a similar project, Nova, at $16/hr — about $5 less than Musen’s rate — after being told Musen would run through year‑end. Mercor says the roles are temporary and disputed some reporting; Meta declined to comment. At its peak Musen employed more than 5,000 people, and Mercor says it manages over 30,000 contractors and pays roughly $1.5M daily, even as some contractors recall prior projects that paid up to $60/hr and company claims of an average $95/hr. The episode is a flashpoint in a wider contraction across the human‑in‑the‑loop labeling economy: firms such as Scale AI, xAI and others have recently cut or reshaped labeling teams, and pay rates are declining as work is consolidated and clients renegotiate terms. Technically, reductions in experienced human raters — who provide critical prompt comparisons, quality labels and nuanced judgments for model training and evaluation — could raise risks to dataset quality, model reliability and long‑tail performance. The situation underscores tensions between cost pressure, workforce precarity, and the centrality of skilled human annotation to current AI model development and safety pipelines.