MIT Tech Review: LLMs form hiring biases more readily than human screeners
MIT Technology Review·high signal
New research covered July 20 finds AI résumé screeners are more likely than humans to develop biases during hiring evaluation — not merely inheriting bias from training data, but generating it in the screening process itself. This matters because résumé pre-screening is one of the most widely deployed production LLM use cases in enterprise HR. Builders shipping any candidate-ranking or scoring feature now have a citable finding that the model is the risk surface, not just the corpus.