Insurance
Home›Insurance›Industry & Deals›Insurers face board-level scrutiny over AI proxy discr…
Insurers face board-level scrutiny over AI proxy discrimination risk
Legal experts warn that even when insurers exclude protected traits, machine learning can still infer them through other correlations used in underwriting.
Insurance Business highlights a governance risk for insurers using machine learning in underwriting, saying algorithms can identify correlations tied to protected characteristics that state law bars for use, including race, gender, and health status.
The article, quoting University of Minnesota Law School professor Daniel Schwarcz, argues the issue is not intent but mathematics, because models that rely on correlations may use other data to proxy for protected traits that predict claims.
It also describes a self-reinforcing competitive dynamic, where insurers may feel pressured to adopt AI because it improves claims prediction, even if regulators permit the use of data that can indirectly discriminate.
The piece points to an evolving regulatory landscape, noting the NAIC adopted a Model Bulletin on insurers’ use of AI in December 2023 and that roughly half of states have adopted it or issued comparable guidance, while testing regulations from prior legislative mandates remain unimplemented in some areas.