Pennsylvania issues AI guidelines for state-licensed insurers
The April 9 release outlines concerns over bias and transparency in AI-supported decisions, particularly in underwriting.
Technology advances much faster than U.S. regulations, and states are scrambling to get ahead of the use of AI in insurance and other industries to ensure these systems make decisions based on true causation, not correlation. On April 9, the Pennsylvania Insurance Department (PID) announced new guidelines for insurers licensed to do business in the state regarding their use of AI. The PID release provides best practices recommendations for how insurers may obtain, develop and implement AI technologies and systems.
“AI may facilitate the development of innovative products, improve consumer interface and service, simplify and automate processes, and promote efficiency and accuracy. However, AI, including AI systems, can present unique risks to consumers, including the potential for inaccuracy, unfair discrimination, data vulnerability, and lack of transparency and explainability. Insurers should take actions to minimize these risks,” reads the PID bulletin.
The PID notice is centered around a model bulletin released December 4, 2023, by the National Association of Insurance Commissioners (NAIC). Pennsylvania is the ninth state to adopt the NAIC model’s recommended AI regulations for insurers, joining Alaska, Colorado, Connecticut, Illinois, Rhode Island, New Hampshire, Nevada and Vermont. Other states have proposed regulations in the works.
The complex regulatory landscape of AI is something executives of insurance companies are keenly aware of, according to Scott Hawkins, head of insurance research and managing director at Conning. He says these regulations also affect vendors as well as the insurers working with them to deploy their technologies.
“The big concern is that they’re not going to be able to deploy the technology the way they would like to maximize the benefit, and if they [insurers] do deploy, they might at some point pay some liability because a new regulation came in,” said Hawkins.
Bias concerns
AI has sped up underwriting processes, allowing insurers to sell more policies, but the technology could have data biases. Regulators raised questions about biases in the risk and pricing models operated with AI, calling for a more transparent process than the standard “black box” decisions. PID and other states that adopted the NAIC model bulletin want insurers to show policyholders and government officials how AI is used and why it came to the conclusions it did for coverage.
“Technology is always evolving and is a great tool to help streamline processes. That said, PID always aims to make certain that insurers are informed on pertinent considerations for using technological advances in a manner that is fair to consumers and is in compliance with current law,” said Pennsylvania Insurance Commissioner Michael Humphreys in the announcement. “AI is no exception. This notice provides insurers with the guidance to help ensure accurate and fair outcomes for Pennsylvanians when using AI.”
The NAIC, PID and other states adopting these AI regulations emphasize the fair and ethical use of AI, holding these systems accountable to state laws and regulations for a safe, secure and fair system with transparency for policyholders and regulators.
Key points
The PID urged insurers licensed in Pennsylvania to review the bulletin to make informed decisions on the company use of AI, ensuring their AI processes do not conflict with established laws and regulations. The key points of the bulletin include:
- PID outlines guidelines for accountability, monitoring, training, audit protocols and governance structures in the AI programs used by Pennsylvania insurers.
- Innovation is encouraged, but insurers must account for risks, such as unfair discrimination, data vulnerability and inaccuracies.
- Insurers should be prepared to provide specific information and documents about their AI systems to PID upon request.
- PID also reminds insurers that AI-supported decisions must comply with applicable insurance laws and regulations.
“The regulatory impact is going to be a major factor in either inhibiting the deployment of these technologies or helping to accelerate them, and it’s an issue that the insurance industry is paying very close attention to,” said Hawkins.
Related: