How ethical AI drives insurance fairness and better models for RGA

Recent advances in artificial intelligence (AI) have captured the imagination of C-suite leaders down to junior staffers, with use cases as diverse as helping diagnose illnesses to planning your next vacation. But how do we know these answers are fair and unbiased? 

For heavily regulated industries such as financial services, this question gains added importance. These services cut to the core of how all of us plot out our lives, fulfill our goals, and protect our loved ones, and the penalties and reputational damage for bias and discrimination can be steep.

When calculating individual risk scores, AI models can draw on a broad range of variables, such as socioeconomic and lifestyle factors, as well as external consumer data and information sources (ECDIS), otherwise known as alternative data, for greater accuracy. But AI often lacks transparency in how it teases out relationships between data points and certain demographics. 

A report by Infopulse highlights the change AI is having within the insurance industry: 77% of leaders in the sector were at some stage of adopting AI across functions in 2024, a jump of 16 percentage points from the prior year.¹ But despite these advances, the prospect of unfair bias and discrimination looms, particularly in the form of unequal pricing and inadequate coverage. As a result, state regulators, such as those in Colorado and New York, are setting up new guardrails around how insurers use AI.

Missouri-based Reinsurance Group of America (RGA), a global reinsurance company that focuses on life and health solutions, leverages AI in its predictive models. The company’s models are built upon vast amounts of data from across the globe, which positions RGA to help carriers better assess risks.

Given the stakes, how could RGA improve confidence with its insurance clients that it has a robust methodology to test for unfair bias? They turned to an advisor with a proven methodology and track record on fairness honed over decades, Ernst & Young LLP (EY US).

EY US would enhance RGA’s approaches for testing the insurance models for compliance, informed by the risks of bias and discrimination, while remaining useful for business purposes. Working together, EY US and RGA established a new bar for responsible innovation.

EY US has a longstanding reputation in the banking industry for discovering and remediating fairness-related issues, particularly in lending, and helping clients comply with regulatory expectations. Our analytical testing is rooted in decades of experience with banking clients on employment law and credit-related areas. Our insurance professionals united those mindsets with a deep background in insurance, including in life insurance modeling, along with the legal and compliance aspects. 

The concept of using AI in insurance is not new, but capabilities and regulations are always evolving. As AI techniques advance, RGA develops new models and updates existing ones on an ongoing basis. In the form of an enhanced testing playbook, EY US delivered a consistent process for RGA to follow for fairness and bias testing at this pivotal moment for AI. 

The playbook’s guidance can apply to any model RGA develops, and it outlines the considerations, decision points, and approaches for how to conduct the testing, which varies by model. It also covers ECDIS, which has drawn the attention of regulators. The testing balances model risk with an assessment of fairness across demographics. This playbook is now a core document for RGA when it develops, tests, and validates models. Model developers, governance teams, and marketing teams all follow a standard process outlined in this playbook.

EY US and RGA team members piloted the testing on two models (one for individual policies and one for group), and RGA will apply the same approach on its other models, both current and future.

In the biannual EY US AI Pulse Survey from late 2025, senior leaders across all sectors whose organizations are investing in AI say that interest in responsible AI has increased over the past year to 67% from 61% a year ago. And more say their focus on ensuring AI operates ethically will increase over the next year (now at 68%, compared with 60% in 2024).

With AI regulatory complexity trending higher, RGA found a way to bring comfort to its clients while even more accurately measuring risk in the life insurance marketplace. The new playbook built by EY US also provides guidelines for applying AI responsibly and confidently as the technology continues to gain powerful new use cases.

“Frameworks and controls are often used as a way to mitigate downside risks and meet regulatory requirements, but they are also the enabler for seizing upside potential,” said Brian Clark of EY US Financial Services. “Equipped with this new playbook, RGA will be able to continue its legacy of innovation in a way that clearly outlines its approach to fairness for their clients’ applicants and policyholders across all demographic backgrounds. They know how to develop better pricing and better underwriting with technology — fairly — to make financial protection more accessible to all and provide people the insurance they need to help them lead longer more fulfilling lives.”