Insurance faces an 'AI proof gap'

Insurance companies are rolling out artificial intelligence faster than they can prove it is safe, fair, and accountable. (grantthornton.com) Grant Thornton said on April 13 that 78% of nearly 1,000 United States business leaders lacked full confidence they could pass an independent artificial intelligence governance audit within 90 days. The same survey found three in four boards had approved major artificial intelligence investments, but only 52% had set clear governance expectations. (grantthornton.com) In insurance, that gap is landing in a regulated business where software can shape underwriting, pricing, claims handling, fraud detection, and customer service. The National Association of Insurance Commissioners adopted its model bulletin on insurers’ use of artificial intelligence on December 4, 2023, and by March 2025, 24 states had adopted it with little or no material change. (content.naic.org) (quarles.com) The model bulletin tells insurers to maintain a written artificial intelligence program with governance, risk management, internal controls, vendor oversight, and consumer notice. Wisconsin’s March 2025 adoption said regulators could ask for documentation covering product development, pricing, claims management, and fraud detection. (quarles.com) Colorado has gone further than guidance. The state adopted amended Regulation 10-1-1 on August 20, 2025, extending governance and risk-management requirements for algorithms and predictive models to individual life, private passenger auto, and health benefit plan insurers, with an effective date of October 15, 2025. (doi.colorado.gov) Courts are also testing how insurers use automated decision tools. In February 2025, a federal judge allowed key claims to proceed in Estate of Gene B. Lokken v. UnitedHealth Group, a class action alleging Medicare Advantage coverage denials were driven by the nH Predict tool rather than individualized medical review. (healthcarefinancenews.com) (litigationtracker.law.georgetown.edu) In a separate case, Cigna was sued in 2023 over its PxDx claims-review system, and a federal judge later allowed class claims to move forward after plaintiffs alleged doctors had denied large volumes of claims without reviewing individual files. Cigna has denied wrongdoing and said the lawsuits do not show improper claim denials. (courthousenews.com) (coregroupusa.com) The industry response is turning into a market of its own. Technology and consulting firms are pitching insurers on “compliant” artificial intelligence stacks, audit trails, and data setups designed to keep models inside approved jurisdictions and produce records for regulators. (cio.com) The immediate question for insurers is no longer whether to use artificial intelligence. It is whether they can show a regulator, a judge, or a customer how a machine-assisted decision was made and who signed off on it. (insurancebusinessmag.com)