NAIC's July 22 Actuarial Panel Maps AI Governance Duties

The NAIC's Big Data and Artificial Intelligence (H) Working Group holds an actuarial panel on AI governance trends on July 22, 2026, paired with a progress update on its 12-state AI Systems Evaluation Tool pilot running through September 2026 (NAIC Big Data and AI Working Group, July 2026). With Fall National Meeting adoption targeted for November, carriers have roughly 90 days after pilot findings close to build exam-ready documentation, and the agenda's framing puts that duty on the signing actuary, not the compliance office alone.

What the Actuarial Panel Is Actually Set to Address

The meeting itself is unremarkable in format: a one-hour Webex session at noon Eastern, open to the public, with two stated purposes on the working group's own calendar entry (NAIC Big Data and AI Working Group, July 2026). The substance is in the pairing. Regulators are not treating "AI governance trends" as a standalone academic panel; they are scheduling it in the same hour as a pilot-program update, which signals the working group wants actuaries hearing both items back to back, connected rather than separate agenda lines.

This is not the group's first actuarial panel on the topic. A comparable session on March 24, 2026 also combined an actuarial governance-trends discussion with an operationalization update on the Model Bulletin (Quarles Law Firm, 2026), suggesting the working group has settled into a recurring format: bring practicing actuaries in front of regulators every few months to describe what governance actually looks like inside a carrier, then measure that description against what the pilot tool is finding in the field. The topics such a panel is built to cover, based on the working group's published charges and the evaluation tool's own structure, are model monitoring cadence, explainability documentation, bias testing protocols, and the escalation thresholds that satisfy the Model Bulletin framework now adopted in 24-plus states and the District of Columbia (Quarles Law Firm, 2026). None of that is new law. It is the operational detail underneath a bulletin regulators adopted in December 2023 and are only now building a standardized way to examine.

The Evaluation Tool's Four Exhibits and the 90-Day Clock

The instrument doing the real work here is the AI Systems Evaluation Tool, piloted across California, Colorado, Connecticut, Florida, Iowa, Louisiana, Maryland, Pennsylvania, Rhode Island, Vermont, Virginia, and Wisconsin from early 2026 through September 2026 (Fenwick, 2026). The tool is built around four exhibits: Exhibit A quantifies a carrier's AI usage across lines and functions, Exhibit B is a governance risk assessment framework, Exhibit C drills into high-risk AI systems specifically, and Exhibit D captures the underlying data details, including a Version 4.0 addition covering reasonable accommodations and policy modifications (Fenwick, 2026). Participating states run monthly coordination calls through the pilot window specifically to avoid sending carriers duplicative, inconsistently worded requests, which is itself a tell that regulators expect this tool to generate real information-gathering friction once it is live everywhere.

The NAIC's own clarification is worth quoting directly, because it defines the boundary of what the tool changes and what it does not: "the tool does not create new requirements for AI governance risk assessments" (NAIC AI Systems Evaluation Tool, Version 4.0, 2026). That single sentence is the entire soft-law posture in miniature. The tool is not a new rule. It is a standardized examination instrument for a rule that already exists, and the mechanics of what each exhibit demands from a working pricing model are detailed separately in our walkthrough of the four-exhibit exam framework.

The published schedule gives the timeline its shape. Pilot data collection runs through September 2026, the tool gets revised based on pilot-state feedback and re-exposed for public comment in September and October, and adoption is expected at the Fall National Meeting in November 2026 (Fenwick, 2026). That sequencing means carriers, particularly those not currently in one of the 12 pilot states, have a real but finite window, roughly the 90 days between the close of pilot data collection and the November adoption vote, to reverse-engineer the exhibit structure against their own AI inventory before it becomes the standard examination request nationally rather than a pilot-state courtesy.

Milestone	Timing	What changes for actuaries
Pilot data collection (12 states)	Early 2026 – September 2026	Exhibits A–D tested against live carrier AI inventories
Tool revision & re-exposure	September – October 2026	Feedback from pilot states incorporated; public comment reopens
Fall National Meeting adoption vote	November 2026	Tool becomes the standardized examination instrument nationally
July 22 actuarial panel	July 22, 2026	Regulators hear directly from practicing actuaries on governance mechanics before the vote

Where the Line Actually Falls: Compliance Officer vs. Signing Actuary

The NAIC's governance expectation, as described in market analysis of the Model Bulletin's implementation, is that a carrier's AI governance committee include actuarial, data science, underwriting, claims, compliance, and legal representation, with senior management or board-level accountability for the program as a whole (Crowell & Moring, 2026). That is a committee structure, not an individual attestation. It answers who sits in the room. It does not answer who signs their name to a specific number.

Precedent for that second question already exists inside actuarial practice, and the July 22 panel is likely to lean on it. The Appointed Actuary issuing a statutory opinion already attests annually to having met the continuing education requirements under Section 3 of the U.S. Qualification Standards before that opinion can stand (Academy Qualification Standards, 2026 cycle). That is a formal, individually signed attestation layered on top of a broader firm compliance program, and it is the closest existing analogue to what an AI-specific actuarial attestation would look like: not a replacement for the carrier's governance committee, but a personal certification that sits on top of it. If a pricing model or a reserve triangle selection is built on an AI-driven output, the compliance officer can own the vendor contract, the training log, and the committee minutes. The signing actuary is the one whose name is on the indication or the opinion when a regulator asks whether the number itself was produced by a properly governed process.

That distinction has a concrete pricing-side illustration already in the record. When a predictive pricing model drifts between filing cycles without triggering a re-validation, the exposure does not sit with whoever approved the original vendor contract; it sits with the actuary who certified the resulting rate indication as still representing the intended risk relationship. The governance committee can document that a model was deployed correctly. Only the actuary can certify that its current output still supports the number in the filing.

Soft Law With Hard Exposure: What the Model Bulletin Does and Does Not Require

The Model Bulletin itself is guidance, adopted by the NAIC in December 2023 and since taken up individually by 24-plus states and the District of Columbia through their own bulletin, regulation, or circular letter processes (Quarles Law Firm, 2026). It carries no statutory penalty provision of its own; a state has to separately incorporate its expectations into examination standards or enforcement action for it to bite. That has already happened in practice. Market conduct examiners in adopting states are using the bulletin's governance expectations, documentation of methodology, data sources, validation results, bias testing, and ongoing monitoring, as the baseline against which they measure a carrier's AI program during routine exams (Crowell & Moring, 2026). That is the mechanism actuaries should sit with: guidance with no direct penalty clause converts into examined-against criteria the moment a state department writes it into its exam procedures, and the Evaluation Tool is precisely that conversion made systematic across 12 states at once. A carrier that treats the bulletin as aspirational because it lacks a fine schedule is misreading how it actually gets enforced.

The Professional Guidance Gap the Panel Cannot Close on Its Own

The regulatory timeline is running well ahead of the profession's own standard-setting process, and that gap is the part trade coverage of the NAIC's AI agenda tends to skip. The SOA Research Institute's Actuarial Intelligence Bulletin posed the open question directly in its January 2026 edition: does the profession need a dedicated AI-specific ASOP, a near-term update to existing standards, or can it rely on current professionalism guidance as-is (SOA Research Institute, January 2026)? The Actuarial Standards Board has not signaled intent to draft a dedicated AI standard, which means the working assumption for now is that ASOP Nos. 56, 23, 41, and 12, covering modeling, data quality, communications, and risk classification respectively, will carry the weight of AI-specific practice through the near term, supplemented by non-binding guidance documents rather than a new enforceable standard.

Separately, the American Academy of Actuaries' Competency Framework began governing new-member requirements as of January 1, 2026, covering general actuarial topics, U.S. laws and practices, and professionalism (American Academy of Actuaries, 2026). It does not yet carve out AI governance as its own competency line distinct from general modeling literacy. Put the two developments side by side and the gap is plain: the professional bodies are still at the stage of asking whether a new standard is warranted, while the regulator is already piloting a four-exhibit examination tool with a November adoption date. Actuaries waiting for an ASOP to tell them how to document AI governance will be waiting past the point at which examiners are already asking for that documentation.

Why This Matters for Signing Actuaries Before November

The practical task for a carrier actuary is to stop treating AI governance documentation as a compliance-department deliverable and start treating it the way reserve support workpapers are treated: something the actuary reviews and can defend personally, because their name, not the compliance officer's, is the one attached to the certified number. Mapping an existing AI inventory against the evaluation tool's Exhibit A through D structure now, regardless of whether the carrier writes business in one of the 12 pilot states, is the single highest-value action available before the Fall vote, since the tool's design is explicitly meant to travel from pilot to national standard with the governance expectations unchanged (NAIC AI Systems Evaluation Tool, Version 4.0, 2026).

The second implication is about timing risk rather than content. Because the profession's own standard-setting apparatus is behind the regulatory calendar, actuaries cannot wait for an ASOP update to define what adequate AI governance documentation looks like for attestation purposes. The Model Bulletin's existing documentation categories, methodology, data sources, validation results, bias testing, and monitoring, are already the de facto content standard being examined against in 24-plus states, ASOP or no ASOP. Building that documentation trail now, in a form a signing actuary would be comfortable personally defending under examination, is the concrete step the July 22 panel is likely to reinforce rather than originate.