Guaranteeing Legal & Regulatory Compliance for AI Models in the Insurance Sector

October 25, 2023

The National Association of Insurance Commissioners (NAIC) recently organized a meeting with its Big data and AI working group. In that meeting they presented and discussed the results of a recent survey about the use of AI applications in the insurance sector. The survey confirmed that AI models are widely used within the sector and reveals that up to 40% of insurance companies are using third-party models.

NAIC and its working groups have been diligently striving to promote the transparent and fair utilization of AI models within insurance companies, guided by five core principles. While these principles are not legally binding, they play a crucial role in informing and establishing the overarching expectations for insurance firms, AI stakeholders, and regulatory bodies, thereby ensuring that AI models maintain the highest standards of ethics, fairness, and safety for consumers. ethical, fair and safe for their clients.

These five core principles are:

Fair & Ethical

With the rapid adoption of AI in the insurance industry, regulators and unions are proactively seeking ways to enhance the trustworthiness of models through ongoing monitoring and the enforcement of compliance measures.

NAIC’s fair & ethical principles acknowledge the inherent risks associated with employing models. In response, insurance companies are expected not only to comply with the sector’s regulations, but also to safeguard and promote consumer interests by implementing standards and controls to ensure the models they employ are free from bias.

Accountable

To ensure transparency and accountability, NAIC emphasizes the insurance sector’s responsibility in ensuring that models produce reliable outcomes. To achieve this, insurance companies are encouraged to maintain comprehensive model documentation that tracks the evolution of AI algorithms, while adhering to insurance laws and regulations in each jurisdiction.

Compliant

NAIC recognizes the significance of both state-wide and federal laws, emphasizing that compliance is an ongoing commitment. It underscores the necessity for all AI systems to consistently align with local, regional, and federal laws.

Transparent

To bolster public confidence, AI stakeholders should make a commitment to safeguard the confidentiality of proprietary algorithms while also delivering comprehensive disclosure of AI systems. This information or documentation should be easy-to-understand and describe the factors that underlie predictions, recommendations, or decisions.

Secure, Safe and Robust

AI/ML algorithms employed by insurance companies are expected to be robust, secure and safe throughout their entire lifecycle. Insurance companies must uphold traceability by maintaining detailed documentation that tracks the evolution of specific models, and they should provide in-depth analysis in accordance with relevant industry best practices and regulatory requirements.

Challenges in NAIC’s principles on Artificial Intelligence

A prominent highlight of these principles is the strong emphasis on insurers maintaining compliance with insurance laws and regulations both on a regional and federal level.
This underscores several challenges in the industry, including:

Dynamic legal landscape

The ever-changing landscape of laws and regulations presents a persistent challenge for insurers. Given the current state of affairs and the swift integration of AI into various business functions, insurance companies can anticipate the introduction of further laws pertaining to AI usage.

State-specific laws

To complicate matters further, insurers must navigate compliance with laws and regulations that significantly vary from one state to another and on the national level. Currently, there are numerous regulations governing AI and model risk, and effectively navigating this intricate web demands additional resources.

Overcoming compliance challenges with Chiron Enterprise

Ensuring regulatory compliance for insurers is a challenge, and adopting advanced solutions like Yields for Governance can be instrumental in achieving and maintaining adherence to ever-changing legal frameworks. How can Yields’ solution, Yields for Performance help you?

Capturing all information

Each legal & regulatory requirement can be translated into
- A number of mandatory controls (either qualitative or quantitative)
- A business process or subprocess
Controls are evidenced as metadata in the inventory – Chiron Enterprise allows you to configure the model inventory so that users are obliged to capture that information
Workflows can be created, updated and modified by configuration only. The process execution serves as an audit trail.

Enforcing compliance

Enforcing compliance through Chiron Enterprise is easy. The platform keeps track of the completeness of all required metadata which it uses as a basis to assess the status of a model and its health throughout its lifecycle. This status can be used to implement quality gates through the deployment pipelines, ensuring that models can only advance to production if they are compliant.

Reporting on compliance

Reporting on compliance is facilitated through standardization and consistency of content elements across models in the inventory. Per state or legislation, Chiron Enterprise enables users to select the required content to populate standard documentation templates, making it possible to generate tailored documentation that also serves as proof of compliance.

Conclusion

Ensuring compliance with constantly evolving regulatory requirements, especially given state-specific variations, is undoubtedly a complex undertaking. Demonstrating compliance necessitates the use of advanced technological tools like Chiron Enterprise to adapt to changing demands.

Stay ahead with expert articles on MRM and AI risk topics, in-depth whitepapers, and Yields company updates.

Streamline your model lifecycle.

Get exclusive insights

Guaranteeing Legal & Regulatory Compliance for AI Models in the Insurance Sector