Regulator outlines potential dangers associated with its opportunities.
The Michigan Department of Insurance and Financial Services (DIFS) has released a bulletin defining regulatory expectations for insurance companies’ development, acquisition, and use of artificial intelligence (AI) systems.
According to DIFS Director Anita Fox, the bulletin defines the types of information that the agency may require when reviewing or examining a carrier’s use of AI.
In an AM Best report, Fox emphasised the need of compliance, adding that while AI is altering multiple industries, including insurance, all choices made by insurers employing these systems must comply with applicable federal and state laws.
Michigan’s advise aligns with New York’s recent legislation to prevent inadvertent discrimination in insurers’ use of AI.
While AI has fostered product innovation and expedited many insurance operations, DIFS has identified possible hazards connected with AI, such as mistakes, discrimination, and difficulties understanding internal processes.
To address these risks, the regulator encouraged carriers to create AI governance frameworks that prioritise transparency, fairness, and accountability while safeguarding trade secrets and private information.
The bulletin suggests that carriers document their AI system validation, auditing, and testing procedures, as well as evaluations of the reliability of AI outputs that may influence decision making.
Additionally, when using third-party AI systems and data, carriers are advised to perform due diligence on potential partners and their data, models, and systems, with documentation of these processes being essential for regulatory investigations.
DIFS stated that, while the bulletin offers guidelines on AI use and the investigation process, it is not thorough. Carriers are free to propose other methods of demonstrating compliance, and this advisory is not meant to prescribe or endorse any one practice or methodology.
What are your opinions about this story? Please feel free to leave your comments below.
