Labeling Pay & Design Gaps
Labeling economics and workforce models remain uneven: reports place Scale AI reviewer pay in the $15–$40+/hr band while programs like Uber AI’s Kenya operation pay per task from roughly $0.50 to $20+. At the same time, companies are recalibrating how they measure employee AI use after recent backtracks and legal scrutiny over recruiting‑platform breaches. Those signals underline that workforce design, pay structure and secure data handling are active inputs into labeling operations today. (x.com) (x.com) (businessinsider.com) (hrdive.com)
The market for training artificial intelligence still runs on human piecework, hourly review jobs and shifting corporate rules about how workers should use the tools they help build. (uber.com) (businessinsider.com) (hrdive.com) Scale AI salary data on Glassdoor showed a “Data Labeler” rate of about $28 an hour, based on employee submissions listed as of October 2025. Uber AI Solutions, which Uber expanded in June 2025, tells workers they can “check out the pay up front” and choose remote gigs task by task. (glassdoor.com) (investor.uber.com) (uber.com) Those two setups describe different labor models for the same basic job: people read, rank, rewrite or tag data so a model learns what “good” output looks like. Uber markets a global task network for labeling, testing and localization, while Scale pitches “reliable AI systems” and hires operators, researchers and engineers directly. (uber.com 1) (uber.com 2) (scale.com) Companies using artificial intelligence inside their own offices are also changing how they measure that work. Business Insider reported on April 13, 2026, that Duolingo Chief Executive Officer Luis von Ahn said the company had “backtracked” on evaluating artificial intelligence use in performance reviews after pushing an “AI-first” approach in April 2025. (businessinsider.com 1) (businessinsider.com 2) That reversal followed a year in which some employers tried to turn artificial intelligence use into a management metric, alongside hiring and head-count decisions. In the April 2025 memo described by Business Insider, von Ahn said Duolingo would use artificial intelligence in hiring, performance reviews and staffing plans; by April 2026, he said some of that push “did not fit.” (businessinsider.com 1) (businessinsider.com 2) At the same time, the systems that move worker data around the hiring pipeline are under legal pressure. HR Dive reported on April 13, 2026, that at least four class-action suits were filed in the Northern District of California after an alleged Mercor breach, with plaintiffs claiming damages tied to exposed personal information. (hrdive.com) Human resources software has already been a litigation target in artificial-intelligence hiring. In August 2025, HR Dive reported that a judge ordered Workday to provide a list of employers that enabled a screening product in a discrimination case tied to automated hiring tools. (hrdive.com) The common thread is operational, not theoretical: who gets paid by the hour, who gets paid by the task, what behavior gets rewarded inside a company and how securely worker data is handled between platforms. Uber says its network offers “specialized experts” and upfront pay visibility; Scale says it builds “reliable AI systems”; Duolingo and recruiting platforms show how quickly those labor systems can be revised once management and legal risks surface. (uber.com) (uber.com) (scale.com) (businessinsider.com) (hrdive.com)
