Insurance MCP governance NAIC Bulletin | Moring
The NAIC Model Bulletin mandates a rigorous, documented Artificial Intelligence System (AIS) program for insurers to mitigate risks in underwriting, pricing, and claims. As adoption spreads across states, compliance leaders must ensure that AI-driven decisions remain transparent, explainable, and non-discriminatory. Proper governance requires proactive oversight of internal models and third-party vendors to satisfy evolving market conduct examination expectations.
Short answer
The NAIC Model Bulletin mandates a rigorous, documented Artificial Intelligence System (AIS) program for insurers to mitigate risks in underwriting, pricing, and claims. As adoption spreads across states, compliance leaders must ensure that AI-driven decisions remain transparent, explainable, and non-discriminatory. Proper governance requires proactive oversight of internal models and third-party vendors to satisfy evolving market conduct examination expectations.
How it works
The NAIC Model Bulletin marks a critical shift in insurance regulation, asserting that the use of Artificial Intelligence Systems (AIS) does not alter an insurer's fundamental legal obligations [4]. For compliance and technology leaders, this means that AI-driven decisions—whether in underwriting, rating, pricing, marketing, or claims—must align with existing regulatory standards [1, 4, 5]. The bulletin is heavily influenced by global principles such as the EU AI Act, emphasizing transparency, accountability, human oversight, and non-discrimination [4]. With roughly 24 states already adopting the model, insurers face a landscape where even non-adopting states may apply these expectations informally through market conduct investigations [3, 4].
Control within this framework centers on the formalization of an AIS program, which requires documented lines of authority connecting AI decisions to senior leadership [3]. This governance structure must ensure that AI systems are not only monitored but also validated and retired according to a clear, written policy [3]. The NAIC requires that the same level of rigor applied to cybersecurity and financial controls be extended to AI [4]. Because the bulletin covers the entire insurance lifecycle, from product development to fraud detection, insurers must maintain visibility into both internal and third-party models to ensure compliance [5].
Regulators are increasingly focused on the intersection of AI and consumer protection, specifically regarding 'Adverse Consumer Outcomes' where an insurer’s decision violates regulatory standards [1]. Sources indicate that regulators view documentation as a primary risk control [5]. When investigating, departments may request evidence of how an insurer approved systems, evaluated model risk, tested for bias, and governed third-party vendors [1, 5]. The expectation is that policies reflect actual practice rather than aspirational goals, as regulators will hold the insurer responsible for compliance failures, including those originating from vendor-supplied platforms [3, 5].
To build a compliant AIS program, practitioners should focus on the following: designate a responsible person or committee for AI oversight; establish a formal, written AIS program that covers the selection, validation, and monitoring of all AI systems; ensure that disparate-impact testing is integrated into the model lifecycle to prevent unfair discrimination; maintain documentation of all model risk evaluations; and implement robust oversight for third-party vendor models and datasets [2, 3, 5]. This checklist serves as the foundation for demonstrating accountability during market conduct exams [3].
Common failure modes include deploying claims-automation AI without sufficient oversight, which can lead to violations of unfair-claims-settlement-practices statutes [2]. Another critical risk is the use of predictive models trained on historical data that inadvertently correlate with protected classes, even when those attributes are not direct inputs [2]. Furthermore, many carriers suffer from a disconnect where they deploy AI tools faster than their claims-compliance function can govern them [2]. Relying on non-adoption as a safe harbor is also a significant risk, as market conduct activity often precedes formal state-level enactment [3].
Teams should begin by assessing their current enterprise risk management and model governance processes to identify gaps in AI oversight [5]. Rather than treating the bulletin as a checkbox exercise, leadership must ensure that the designated governance function has actual visibility into which AI systems are active and what decisions they influence [3]. By integrating disparate-impact testing and explainability requirements directly into the model development lifecycle, insurers can move from reactive compliance to a proactive stance that satisfies NAIC-aligned expectations [2].
Frequently asked questions
How does the NAIC Bulletin impact the use of third-party AI models in insurance?
The NAIC Bulletin makes it clear that insurers are responsible for the compliance of all AI systems influencing insurance decisions, regardless of whether they are developed internally or provided by a third-party vendor. Effective MCP governance requires that insurers perform due diligence on vendor-supplied models, ensuring they are validated, monitored, and tested for bias just as strictly as internal systems to prevent regulatory non-compliance.
What role does documentation play in an insurance AIS program?
Documentation serves as the primary evidence for regulators during market conduct exams to prove that an insurer is managing AI risk appropriately. Under the NAIC framework, insurers must maintain written records of how they approve, validate, monitor, and retire AI systems. Strong MCP governance relies on this documentation to demonstrate that AI-driven insurance decisions are explainable, reasonable, and consistently applied across the organization.
Why is disparate-impact testing critical for insurance pricing models?
Predictive models often inherit historical patterns that may correlate with protected classes, even when those attributes are not used as inputs. To satisfy NAIC expectations, insurers must build disparate-impact testing into their model lifecycle to prevent unfair discrimination. Proper MCP governance ensures that pricing models are tested for bias before deployment, allowing the insurer to defend its underwriting decisions to regulators during an examination.
How should insurers structure their AI oversight committee?
The NAIC Bulletin requires the designation of a responsible person or committee to oversee AI use, with clear lines of authority connecting AI decisions to senior leadership. This function must have actual visibility into active AI systems and their impact on insurance operations. Effective MCP governance ensures this committee is not a checkbox role but an active oversight body capable of managing model risk and compliance.