Liberty Mutual uses machine learning models trained on over 200 million data points and 5 million historical claims to predict claim outcomes, flag potentially fraudulent activity, and guide claims teams toward better decisions. The company's official AI use notice confirms fraud detection and claims processing as active AI applications.

Liberty Mutual's claims AI applies natural language processing and predictive modeling to unstructured data including images, notes, calls, and records to identify high-value or high-risk claims early. The company has reported that its automated fraud model is 20 times better at identifying fraudulent activity than random chance, and that it can predict claims with compensability issues at a 60% rate. These models feed insights to human claims managers rather than making autonomous final decisions. The $25 million, five-year partnership with MIT's Stephen A. Schwarzman College of Computing, announced around 2019, was established in part to advance AI research in areas such as computer vision, natural language processing, and risk-aware decision-making relevant to claims.