The Insured Retirement Institute (IRI) said modifications should be considered for a proposed model from the National Association of Insurance Commissioners (NIAC) on insurers’ use of algorithms, predictive models, and artificial intelligence (AI).
IRI officials said they share the NAIC’s goal of ensuring that decisions impacting consumers that are made or supported by advanced analytical and computational technologies comply with insurance laws and regulations.
However, in comments submitted to the NAIC Innovation, Cybersecurity and Technology (H) Committee, IRI said its members have concerns around the expectations for third-party AI systems and contracts with such third parties.
“Third-party vendors may be unwilling to provide proprietary information regarding their data or models directly to insurance companies, and we believe recognition of this issue within the Bulletin is important,” Sarah Wood, director of state policy and regulatory affairs at IRI, said.
Wood added that the separate risks for the different types of AI technologies should be addressed.
“We believe that there is a difference between predictive models that are trained on defined, labeled data (supervised by humans) and AI technologies that cannot be fully supervised (i.e., ChatGPT) and may or may not lead to a specific outcome,” Wood wrote in the comments. “It would be appropriate to differentiate between these different models and create a standard that is separate for each modeling type. We’d be happy to collaborate further with the Committee on how to address this issue.”