The rules in 2026

What Are the AI Rules for Insurance Agencies in 2026?

Updated November 11, 2026

Short answer: No single AI law governs insurance agencies. The rules that matter are the data rules you already live under: the NAIC Insurance Data Security Model Law (#668), which reaches licensed agencies in the states that adopted it; the state safeguards rules built on GLBA and the NAIC's #673; and, in New York, the DFS cybersecurity regulation. The NAIC's AI bulletin and Colorado's SB21-169 get the headlines, but both bind insurers, not agents. This page takes each one in turn and ends with what applies to a 12-person agency.

The NAIC Insurance Data Security Model Law (#668)

The NAIC adopted #668 in the fourth quarter of 2017 and made a technical edit in 2025 (NAIC #668). It is a model law, so it has no force until a state enacts it, and the version your state passed is the one to read.

It applies to agencies. #668 is written for any "Licensee", which it defines as "any Person licensed, authorized to operate, or registered, or required to be licensed, authorized, or registered pursuant to the insurance laws of this State". A licensed agency or producer is a licensee.

What it asks of a licensee:

  • A written program, scaled to the business. "Commensurate with the size and complexity of the Licensee, the nature and scope of the Licensee's activities, including its use of Third-Party Service Providers ... each Licensee shall develop, implement, and maintain a comprehensive written Information Security Program based on the Licensee's Risk Assessment" (section 4A).
  • Oversight of vendors. A licensee "shall exercise due diligence in selecting its Third-Party Service Provider" and "shall require a Third-Party Service Provider to implement appropriate administrative, technical, and physical measures" to protect the information it holds (section 4F). An AI vendor that handles client data fits the law's definition of a third-party service provider.
  • Notice within 72 hours. A licensee must "notify the Commissioner as promptly as possible but in no event later than 72 hours from a determination that a Cybersecurity Event has occurred" (section 6A). For an agency, that applies to its home-state commissioner, and to another state's commissioner when the event involves "250 or more Consumers residing in this State" and meets further conditions.
  • A vendor's breach is handled as the agency's own. For an event at a third-party service provider, "the Licensee shall treat such event as it would under Section 6A" (section 6D). Under #668 your clock starts the day after the vendor tells you, or the day after you otherwise learn of the event, whichever is sooner.

Other points agencies ask about:

  • Insurers must tell you about breaches that touch your clients. "the insurer shall notify the producers of record of all affected Consumers as soon as practicable as directed by the Commissioner" (section 6F).
  • The annual February 15 certification is for insurers. It applies to "each insurer domiciled in this State" (section 4I), not to agencies.
  • There are exemptions for small agencies and for agents covered by someone else's program. They are narrower than many people think; the small agency page sets them out.
  • New York compliance was meant to count. A drafting note says the drafters "intend" that a licensee in compliance with 23 NYCRR 500 "is also in compliance with this Act". Check whether your state's version says so.
  • No private lawsuits. The model creates no private right of action (section 2B).

Which states have adopted #668?

The NAIC's state page for #668 (Summer 2026 edition) lists these jurisdictions with a citation in its "Model Adoption" column (NAIC state page, PDF):

Alabama, Alaska, Connecticut, Delaware, Hawaii, Illinois, Indiana, Iowa, Kentucky, Louisiana, Maine, Maryland, Michigan, Minnesota, Mississippi, Missouri, New Hampshire, North Dakota, Ohio, Oklahoma, Pennsylvania, Puerto Rico, Rhode Island, South Carolina, Tennessee (the chart notes "portions of model"), Vermont, Virginia and Wisconsin.

Several large states are not on that list. The chart puts California, Colorado, New York and Texas under "Related Activity" instead, meaning they have rules on the same subject that the NAIC does not count as adopting the model. New York's entry is its cybersecurity regulation, covered below.

The chart carries its own warning, and it applies here too: "This chart does not constitute a formal legal opinion by the NAIC staff ... Readers should consult state law for further details and for the most current information."

GLBA and the state safeguards rules (#673)

GLBA puts a duty on every financial institution "to protect the security and confidentiality of those customers' nonpublic personal information" (15 U.S.C. 6801). For anyone "engaged in providing insurance", the state insurance authority enforces it (15 U.S.C. 6805(a)(6)), and the states must put the safeguards standards in place "by rule". Our reading is that this leaves a licensed agency under its state insurance safeguards rule, not the FTC's Safeguards Rule.

The NAIC's template for those state rules is the Standards for Safeguarding Customer Information Model Regulation (#673), adopted in 2002 (NAIC #673). It asks for "a comprehensive written information security program ... appropriate to the size and complexity of the licensee" (section 3). As its example of how to run that program, section 8 gives a licensee that "Exercises appropriate due diligence in selecting its service providers" and "Requires its service providers to implement appropriate measures designed to meet the objectives of this regulation". The model calls sections 6 to 9 "examples of methods of implementation", so check how your state's version words them. It covers health information as well as financial information.

Our reading of #673's text is that it has no exemption for agencies with fewer than 10 employees, unlike #668. Ask your counsel how your state's version reads.

The NAIC AI bulletin applies to insurers, not agencies

The NAIC's Model Bulletin on the Use of Artificial Intelligence Systems by Insurers was adopted on December 4, 2023 (NAIC bulletin, PDF). It is addressed "TO: All Insurers Licensed to Do Business In" the adopting state. The words "producer" and "agent" do not appear in its text (our search of the PDF). It does not regulate your agency's use of a chat assistant directly.

What it asks of insurers:

  • A written AI program. Insurers "are expected to develop, implement, and maintain a written program (an 'AIS Program') for the responsible use of AI Systems that make, or support decisions related to regulated insurance practices." The program "should be tailored to and proportionate with the Insurer's use and reliance on AI and AI Systems."
  • Generative AI is in scope. The bulletin defines it as "a class of AI Systems that generate content in the form of data, text, images, sounds, or video".
  • Oversight of third parties. Insurers are to cover "Due diligence and the methods employed by the Insurer to assess the third party and its data or AI Systems", and, "Where appropriate and available", contract terms that "Provide audit rights" and "Require the third party to cooperate with the Insurer with regard to regulatory inquiries." A "Third Party" is "an organization other than the Insurer that provides services, data, or other resources related to AI."
  • It restates existing law. The bulletin is guidance, issued "to remind all Insurers" that they "must comply with all applicable insurance laws and regulations."

So where does an agency come in? Our reading is that an agency is touched mainly in two ways: if it provides AI-related services to an insurer, it may be that insurer's "Third Party", and a carrier may pass some of its expectations to its agencies through their agreements. Read your carrier agreements for AI and data terms.

Where the bulletin has been adopted. The NAIC's map (status as of April 1, 2026) lists these as "Adopted States": Alaska, Arkansas, Connecticut, Delaware, District of Columbia, Hawaii, Illinois, Iowa, Kentucky, Maryland, Massachusetts, Michigan, Nebraska, Nevada, New Hampshire, New Jersey, North Carolina, Oklahoma, Pennsylvania, Rhode Island, Vermont, Virginia, Washington, West Virginia and Wisconsin (NAIC map, PDF). The map says it "does not reflect a determination as to whether the pending or enacted legislation contains all elements of the model." It also lists insurance AI guidance that is not the model: California Bulletin 2022-5, Colorado's 3 CCR 702-10, New York Circular Letter No. 7 (July 11, 2024) and Texas Bulletin B-0036-20. The map is several months old, so check it for later adoptions.

New York: the DFS cybersecurity regulation (23 NYCRR 500)

New York agencies work under the Department of Financial Services' cybersecurity regulation, 23 NYCRR 500. The DFS's exemptions page explains who is exempt, and the regulation's text is in the DFS's official copy of Part 500; we read both on November 11, 2026. The regulation says:

  • Vendor policies. Section 500.11 asks each covered entity to "implement written policies and procedures designed to ensure the security of information systems and nonpublic information that are accessible to, or held by, third-party service providers", including "notice to be provided to the covered entity in the event of a cybersecurity event directly impacting ... the covered entity's nonpublic information being held by the third-party service provider" (23 NYCRR 500). The section was last amended effective November 1, 2023.
  • 72-hour notice, including incidents at vendors. Section 500.17(a)(1) asks for notice "as promptly as possible but in no event later than 72 hours after determining that a cybersecurity incident has occurred at the covered entity, its affiliates, or a third-party service provider".
  • An annual filing by April 15. Covered entities submit either a certification or an acknowledgment each year.

The DFS exemptions page sets out two levels of exemption: entities with a full exemption "do not need to comply with any requirements", and those with a limited exemption "need to comply with some requirements". Small agencies usually look at the limited exemption. The small agency page lists its three tests and what it still requires, including the vendor policy in 500.11.

Colorado: SB21-169 binds insurers, not agents

Colorado's SB21-169 was signed on July 6, 2021. According to the Colorado Division of Insurance, it "holds insurers accountable for testing their big data systems - including external consumer data and information sources, algorithms, and predictive models - to ensure they are not unfairly discriminating against consumers on the basis of a protected class" (Colorado DOI). Agents appear only as stakeholders the Commissioner works with "before adopting rules". The duties fall on insurers, so SB21-169 does not directly regulate an agency's use of a chat assistant.

The rules under it are in Regulation 10-1-1 (3 CCR 702-10). The amended regulation covers life insurers, private passenger auto insurers and health benefit plan insurers, and took effect on October 15, 2025. The NAIC's map gives the original regulation's effective date as November 13, 2023. Insurers that do not use external consumer data and information sources file an attestation by December 1 each year. The Division's page shows a separate quantitative testing regulation as a draft, released for comment on September 28, 2023.

What applies to a 12-person agency?

Take an independent agency with 12 people, including its independent contractors, that uses an AI tool on client information. This is our reading of the rules above, not legal advice:

Rule Does it apply? What it means for the AI tool
#668, in a state that adopted it Yes. At 12 people the fewer-than-10 exemption does not apply. A written information security program that covers the AI vendor, due diligence before you choose it, safeguards required of it, and the 72-hour notice path
The state's GLBA safeguards rule, if it follows #673 Yes. Our reading is that #673 has no small-agency exemption. A written program sized to the agency, with vendor due diligence as #673's example of how to run it
23 NYCRR 500, for a New York agency Yes, but under 20 people the agency may qualify for the limited exemption. Even then, a written third-party service provider policy (500.11), a risk assessment, multi-factor authentication and the notice duties
The NAIC AI bulletin Not directly. It is addressed to insurers. Read carrier agreements for AI and data terms the carrier passes on
Colorado SB21-169 and Regulation 10-1-1 Not directly. It binds insurers. None for the agency's own chat tool

The common thread is the vendor. Whatever AI tool the agency picks, the rules that do apply treat it as a service provider holding client information, which is why the vendor due diligence page is the next place to go.

This guide is general information, not legal or compliance advice. Model laws and charts are a starting point; your own state's statute and regulator control.

About Faradex. Faradex is a private AI workspace for small and mid-sized firms that handle confidential client work. Each firm gets its own Workspace on dedicated infrastructure in the United States. Files and conversations are held in memory and are gone when the Workspace is reset. AI model: Currently Opus via AWS Bedrock, more information is available on our website. AWS does not store prompts or responses, and model providers have no access to them. Pricing: from $120 a month, no seat minimum; see faradex.ai/pricing.