← Back to feed

Harm Laundering in GPT Models: Gender Discrimination Transformed Rather Than Eliminated

AI · Studies · 2026-09-19 04:07

An arXiv paper reveals that gender bias in GPT models is not eliminated but transformed and concealed through a mechanism the authors call "harm laundering," exposing a new dimension of bias in large language models.

Source: arxiv.org

Auto-generated summary for reference only. RSS