NAIC sets 2026 pilot to test insurers' AI use as trade groups push back
State insurance regulators are moving ahead with a 2026 pilot to test an AI Systems Evaluation Tool, despite sharp criticism from industry trade groups. Ten insurers are slated to participate, according to Iowa Insurance Commissioner Doug Ommen, who chaired the Big Data and Artificial Intelligence Working Group meeting in Florida. The pilot's structure is still being defined.
After the pilot, regulators plan an evaluation period to refine the tool. "At the conclusion of the pilot period, we'll then hear from the pilot group and consider lessons learned," Ommen said. He emphasized the pilot will be instructive for both regulators and participating carriers.
What the tool is - and how it may be used
The AI Systems Evaluation Tool was proposed this summer as an option for market conduct exams. Regulators are not required to use it, but it gives exam teams a standardized way to review how carriers design, validate, deploy, and monitor AI systems across the value chain.
Insurers have moved from basic automation to advanced AI, including generative models for customer engagement and internal operations. Regulators continue to seek guardrails that keep pace with these changes while protecting consumers through fairness, transparency, and accountability.
Federal backdrop and state authority
Ommen acknowledged federal efforts that could preempt state law through a single national AI framework. He expressed confidence in state regulators' authority to coordinate supervision of AI in insurance. "Any attempt to prevent us from doing our work here would impact consumers negatively and represent a significant departure from our state regulatory system that's worked for over 150 years," he said.
For background on state insurance perspectives on AI, see the NAIC's AI resources page at NAIC CIPR: Artificial Intelligence.
Existing guidance and parallel state efforts
In December 2023, the NAIC adopted a Model Bulletin on the use of algorithms, predictive models, and AI systems by insurers. It is guidance rather than a model law or regulation, intended to align AI use with existing market conduct, corporate governance, and unfair and deceptive practice laws.
Several states have proceeded with their own rules. Meanwhile, NCOIL advanced work on a Model Act regarding insurers' use of AI, which is not yet final. Learn more at NCOIL.
Industry concerns with the pilot
During a line-by-line review of the Nov. 19 second draft, trade associations representing life, health, P&C, mutual, and reinsurance carriers submitted a joint Dec. 5 letter citing major issues. Their core concerns include scope, timing, and the potential enforcement posture during a pilot that is still in draft form.
- Voluntary for regulators but compulsory for companies
- No defined pilot duration
- Tool may be used in financial and/or market conduct exams
- Risk of penalties tied to "negative" findings during the pilot phase
- Pilot could begin before the final tool is exposed for public comment
The letter urges the NAIC to mirror prior pilot processes: make company participation voluntary and use gathered information solely to develop the tool, not for compliance actions.
What carriers should do now
- Build an AI system inventory: catalog use cases across underwriting, pricing, claims, distribution, and service. Identify owners, model types, data sources, and consumers impacted.
- Tighten governance: assign accountable executives, define approval gates, and establish escalation paths for model risk, bias, and consumer harm.
- Document rigorously: data provenance and rights, feature selection, validation plans, monitoring thresholds, explainability methods, and human-in-the-loop controls.
- Test for fairness: run disparate impact analyses where applicable, stress test edge cases, and require vendors to supply test results and remediation plans.
- Clarify use constraints: spell out acceptable and prohibited AI uses for underwriting, claims, SIU, marketing, and customer service. Align notices and adverse action communications.
- Strengthen model lifecycle controls: versioning, change management, rollback plans, and auditable logs covering training, fine-tuning, and prompt or parameter changes.
- Manage third parties: embed audit rights, performance SLAs, data handling rules, and bias remediation obligations into contracts. Verify sub-processor exposure.
- Protect data: minimize sensitive attributes, enforce retention limits, and secure PHI/PII with encryption and access controls. Track synthetic data and enrichment vendors.
- Prepare for incidents: define kill switches, contingency procedures, customer remediation steps, and regulator notification playbooks.
- Run a mock exam: practice the tool's likely requests, close documentation gaps, and train front-line teams, compliance, audit, and model owners.
Key watch items
- Ten insurers expected in the 2026 pilot; structure still in development
- Post-pilot evaluation to refine the tool
- Potential use in market conduct and financial exams
- Pilot duration and enforcement posture not yet defined
- Another draft of the tool and public comment timing
If your teams need structured, role-based upskilling on safe and compliant AI use, explore practical programs here: AI certifications and pathways.
Your membership also unlocks:
