States, Feds Clash Over AI in Insurance
States on both sides of the aisle are moving to limit how insurers use AI in claims and underwriting, citing concerns over algorithmic bias and data privacy. The push for state-level control is creating a complex regulatory patchwork, with the federal government potentially looking to override those limits, adding significant uncertainty for InsurTechs.
Colorado is a key state to watch, having enacted SB 21-169, a first-of-its-kind law prohibiting insurers from using external data and AI models in a way that results in "unfair discrimination" based on protected characteristics. The law requires life, and now health and auto, insurers to establish a risk-management framework, conduct annual bias reviews, and submit compliance reports to the state's Division of Insurance. The National Association of Insurance Commissioners (NAIC) has stepped in to create a national standard, adopting a model bulletin on AI use in late 2023. This non-binding guidance, adopted by at least 24 states, requires insurers to develop a written program for responsible AI use and to test models for bias. The NAIC is now considering drafting a formal model law to create more uniform statutory requirements across all states. The push for federal preemption centers on the idea of a single, uniform regulatory framework rather than a patchwork of state laws. This debate isn't new; the McCarran-Ferguson Act of 1945 generally affirmed state-level regulation of the insurance industry. A recent executive order, however, has directed the Attorney General to establish a task force to challenge state AI laws, explicitly citing Colorado's as potentially problematic. At the heart of the debate is the risk of "algorithmic bias," where AI models perpetuate and even amplify historical biases found in training data. For example, an AI underwriting tool might use seemingly neutral data like ZIP codes or credit scores, which can act as proxies for race or ethnicity, leading to discriminatory outcomes in pricing or coverage denial. In response to regulatory pressure, insurers are being pushed to ensure a "human in the loop" for critical decisions, particularly for claim denials. Florida's Insurance Commissioner, Mike Yaworsky, has advocated for policies that ensure human oversight and require consumer notification when AI is used in decision-making processes. The insurance industry remains divided on the path forward. While some executives see AI as a top strategic initiative for improving efficiency and risk assessment, there is significant concern about regulatory uncertainty and the potential for a consumer trust backlash. The NAIC itself has noted division among its members on the necessity of a binding model law.
