New Research Shows AI Systems Develop Their Own Biases in Hiring Decisions Beyond Training Data

MIT researchers have found that large language models used in recruitment screening not only inherit human biases from their training data but also independently generate new biases during the hiring process. The findings raise concerns about the fairness of AI-driven résumé screening, which is increasingly used by employers as a first-pass filter before human review.

Why it matters: As AI hiring tools become more prevalent, understanding how these systems develop and amplify bias is critical for AI practitioners, marketers, and HR leaders relying on automated candidate evaluation.