Building Regulatory Trust in New Underwriting Technologies

The rapid integration of artificial intelligence and machine learning into insurance underwriting has created a critical challenge for the industry: building regulatory trust in new underwriting technology. As carriers increasingly rely on complex algorithms and vast datasets to assess risk, they must also navigate a fragmented and evolving regulatory landscape. The core issue is no longer if carriers will adopt these tools, but how they can do so in a way that is transparent, fair, and compliant in the eyes of state and federal oversight bodies.

"Insurers are responsible for the use of AI systems, even if that use is by a third-party." - National Association of Insurance Commissioners (NAIC), Model Bulletin on the Use of Artificial Intelligence Systems by Insurers (2023)

The deep dive into algorithmic underwriting and regulatory trust

The shift from manual, human-driven underwriting to automated, AI-powered systems represents a fundamental change in how risk is evaluated and priced. This is where building regulatory trust in new underwriting technology becomes a central pillar of a modern compliance strategy. Regulators are no longer just auditing rulebooks and policy documents; they are now scrutinizing the data, models, and governance frameworks that underpin automated underwriting decisions. The "black box" nature of some sophisticated models presents a significant hurdle. When an insurer cannot fully explain how a particular decision was reached, it raises immediate red flags for regulators focused on preventing unfair or discriminatory outcomes.

The NAIC has been proactive in this area, establishing the Big Data and Artificial Intelligence (H) Working Group in 2019 to study the use of AI in insurance. This work culminated in the adoption of the Model Bulletin on the Use of Artificial Intelligence Systems by Insurers in late 2023. This bulletin provides crucial guidance for carriers, emphasizing that insurers retain ultimate responsibility for the outcomes produced by their AI systems, regardless of whether they were developed in-house or by a third-party vendor. This highlights the need for robust due diligence and ongoing monitoring of all algorithmic underwriting tools.

Feature	Traditional Underwriting	AI-Powered Underwriting
Primary Data Sources	Applications, medical exams, MIB	EHR, wearables, public records, digital biomarkers
Decision Speed	Days or weeks	Seconds or minutes
Process Transparency	Manual review with clear rationale	Can be opaque ("black box") without proper documentation
Scalability	Limited by human resources	Highly scalable
Regulatory Oversight	Established and well-understood	Evolving; focused on fairness and model governance

Key principles from the NAIC's guidance on AI are critical for carriers to understand and implement:

• Fairness and Non-Discrimination: Insurers must be able to demonstrate that their AI models do not result in unfair discrimination against protected classes.
• Transparency and Explainability: Carriers need to maintain documentation that allows for the reconstruction of how their AI models make decisions.
• Governance and Accountability: A formal, written program should be in place that establishes clear lines of responsibility for the AI system's entire lifecycle.
• Risk Management: Insurers are expected to have a robust framework for identifying, measuring, and mitigating risks associated with their use of AI.

Industry applications and compliance frameworks

To build regulatory trust, carriers must move beyond high-level policy statements and implement concrete, auditable compliance frameworks for their new underwriting technologies. This involves a multi-faceted approach that integrates legal, data science, and operational teams.

Robust data governance

The foundation of a trustworthy AI underwriting system is a strong data governance program. This includes clear policies for data quality, data lineage, and data privacy. For every data point used in a model, a carrier should be able to answer: Where did it come from? How is it secured? Is its use permissible for this specific underwriting purpose?

Model risk management

A formal model risk management (MRM) program is essential. This framework should cover the entire lifecycle of a model, from development and validation to deployment and ongoing monitoring. Key components of a strong MRM program include:

• Independent Validation: Models should be validated by a team that is separate from the development team to ensure an unbiased assessment of their performance and fairness.
• Bias Testing: Rigorous testing for potential bias against protected classes is a regulatory expectation. This should be done before a model is deployed and on an ongoing basis.
• Drift Monitoring: Models can become less accurate over time as the underlying data changes. Continuous monitoring for model drift is crucial to ensure ongoing accuracy and fairness.

Consumer disclosures and recourse

Transparency extends to the consumer. Carriers must be able to provide consumers with a clear explanation of adverse decisions. This requires moving beyond generic, high-level reason codes and providing specific, accurate information about the data that influenced the decision. Furthermore, a clear process for consumers to question and correct inaccurate data is a critical component of a fair system.

Current research and evidence

The academic and research community is actively engaged in studying the implications of AI in underwriting. A 2022 report from researchers at the Stanford Institute for Human-Centered Artificial Intelligence highlighted the challenges of ensuring fairness in algorithmic systems when historical data reflects societal biases. Similarly, research from the Casualty Actuarial Society has explored various techniques for testing for and mitigating bias in pricing models. The NAIC's own research, conducted through its working groups, provides an invaluable resource for carriers, offering insights into regulatory expectations and best practices. The 2023 adoption of the AI Model Bulletin was a direct result of years of this focused research and collaboration.

The future of regulatory technology in underwriting

The future of building regulatory trust in new underwriting technology lies in greater standardization and the emergence of specialized "RegTech" solutions. We can expect to see the development of more sophisticated tools for automated model monitoring and auditing. The NAIC is already exploring the concept of a third-party data and models regulatory framework, which could create a more streamlined process for validating external data and models. This would allow regulators to develop a deeper understanding of these tools and provide greater certainty for insurers. The ultimate goal is to create an ecosystem where innovation can flourish within a framework of robust regulatory oversight, ensuring that new technologies are used responsibly and for the benefit of all consumers.

Frequently asked questions

What is the most significant challenge in building regulatory trust for new underwriting technologies?

The biggest challenge is the "black box" problem. Many advanced AI models are so complex that it is difficult to explain the exact reason for a specific decision. This lack of transparency is a major concern for regulators who are focused on ensuring fairness and preventing discrimination. Carriers must invest in explainable AI (XAI) techniques and robust documentation to overcome this hurdle.

How does the NAIC's AI Model Bulletin affect carriers?

The NAIC's Model Bulletin provides a clear set of expectations for insurers using AI. It emphasizes that insurers are ultimately responsible for the outcomes of their AI systems, even if they use a third-party vendor. It requires carriers to have a written AI program, conduct risk management, and ensure their models are fair and transparent. While it is a model bulletin, it is expected to be widely adopted by state regulators.

What is model drift and why is it a regulatory concern?

Model drift occurs when the performance of an AI model degrades over time due to changes in the underlying data. This is a regulatory concern because a model that was initially fair and accurate could become discriminatory or inaccurate as the data it processes changes. Regulators expect carriers to have systems in place to monitor for model drift and to retrain or replace models when necessary.

As the regulatory landscape for digital underwriting continues to mature, Circadify is at the forefront of developing solutions that help carriers navigate these complex requirements. Our platform is designed with compliance at its core, enabling insurers to build and maintain regulatory trust as they adopt new technologies. To learn more about how to create a compliance-first approach to digital transformation, explore our compliance guides and regulatory insights at circadify.com/industries/payers-insurance.