Can insurance agents use ChatGPT? What state insurance rules say about client files

The Morse teamPublished November 6, 202610 min read

Yes, insurance agents can use ChatGPT for work with no client in it, but a client's file belongs only in an AI tool your agency has vetted and bound by contract. As of November 2026, a client's Social Security number, driver's license number, account number, or health details count as "nonpublic information" under the insurance data security laws of 28 jurisdictions and New York's cybersecurity rule, and each of those laws treats an AI company that receives them as a third-party service provider.

This isn't legal advice. It's about where client information goes when you paste it into AI, and which rules follow it there. On November 6, 2026, we read the Gramm-Leach-Bliley Act's enforcement section, the NAIC's Insurance Data Security Model Law and its AI bulletin with the NAIC's adoption maps, New York's Part 500 and three of its Department of Financial Services (DFS) letters, HIPAA's rules, the Driver's Privacy Protection Act, the Fair Credit Reporting Act, a carrier agency contract review from the Independent Insurance Agents & Brokers of America (the Big I), and OpenAI's help pages. New York's DFS deserves credit here. Its 2024 AI letter tells regulated firms, agents included, exactly how to keep client data out of public AI tools.

Does the FTC Safeguards Rule apply to insurance agents?

Generally, no. For insurance, your state insurance department enforces federal financial privacy law instead. The Gramm-Leach-Bliley Act hands enforcement "in the case of any person engaged in providing insurance" to "the applicable State insurance authority," and leaves the Federal Trade Commission only financial institutions no other regulator covers (15 U.S.C. 6805). The FTC's own guide says "The Safeguards Rule applies to financial institutions subject to the FTC's jurisdiction," and excludes those overseen by another regulator under that section (FTC). Guides written for agencies still treat the FTC rule as an open question. One, from July 2023, says it "may or not apply to your Agency" (Kite Technology Group). The rule does name tax preparers, which we cover in can accountants use ChatGPT.

That moves the question to your state. The same act requires each state insurance regulator to set safeguards standards for the firms it oversees (15 U.S.C. 6801).

Which states have adopted the NAIC Insurance Data Security Model Law?

As of November 3, 2026, the NAIC's map shows 28 jurisdictions with the model or similar laws: Alabama, Alaska, Connecticut, Delaware, Hawaii, Illinois, Indiana, Iowa, Kentucky, Louisiana, Maine, Maryland, Michigan, Minnesota, Mississippi, Missouri, New Hampshire, North Dakota, Ohio, Oklahoma, Pennsylvania, Rhode Island, South Carolina, Tennessee, Vermont, Virginia, Wisconsin, and Puerto Rico. Washington, D.C. is pending, and New York has its own rule (NAIC). California, Texas, Florida, Georgia, New Jersey, and North Carolina are not on the list, so agents there look to their state's other privacy rules and their carriers' contracts.

Where it's law, the model reaches you directly. A "Licensee" is anyone "licensed, authorized to operate, or registered" under the state's insurance laws, which includes producers (NAIC Model 668). Its nonpublic information covers a consumer's name or other identifier combined with a "Social Security number," a "Driver's license number," an "Account number, credit or debit card number," or "Biometric records," plus nearly any health information "except age or gender." An AI company that processes or stores that information for you is a "Third-Party Service Provider," and the model says you "shall exercise due diligence in selecting" one and "shall require" it to "implement appropriate administrative, technical, and physical measures" to protect the data. Some cybersecurity events have to be reported to the commissioner within 72 hours of determining they happened.

Small agencies get a partial pass in the model text. A licensee "with fewer than ten employees, including any independent contractors, is exempt from Section 4," the section that holds the vendor rule. The 72-hour notice sits in Section 6 and still applies. The NAIC's map also says it doesn't show whether a state's law "contains all elements of the model," so check your own statute.

What does New York's Part 500 say about agents using AI?

It covers you, and its small-agency exemption keeps the vendor rule. Part 500 applies to "any person operating under or required to operate under a license" under the Insurance Law, and its definition of nonpublic information closely tracks the NAIC model's (23 NYCRR 500.1). An agency with fewer than 20 employees and contractors, less than $7.5 million in revenue in each of the last three years, or less than $15 million in assets gets a limited exemption, but DFS lists the third-party service provider policy in 500.11 among the sections it must still follow (DFS). That policy covers "due diligence processes" for each provider and its "use of encryption" (23 NYCRR 500.11).

DFS has applied this to AI twice. Its October 16, 2024 letter says staff allowed to use AI should be trained "on how to draft queries to avoid disclosing NPI," and suggests blocking queries that "might expose NPI to a public AI product or system." It says it "does not impose any new requirements" beyond Part 500 (DFS). Its October 21, 2025 letter on vendors says a regulated firm "may not delegate responsibility for compliance" to a provider, and suggests contracts address "whether the Covered Entity's data may be used to train AI models" and "where data may be stored, processed, or accessed" (DFS).

DFS's Circular Letter No. 7 of 2024, on AI in underwriting and pricing, is addressed to "All Insurers Authorized to Write Insurance in New York State" (DFS). It doesn't mention agents or brokers.

Does the NAIC AI bulletin apply to insurance agents?

Not directly. The NAIC adopted its Model Bulletin on the Use of Artificial Intelligence Systems by Insurers on December 4, 2023, and it is addressed "TO: All Insurers Licensed to Do Business" in the adopting state (NAIC). As of November 3, 2026, 29 jurisdictions have adopted it, most recently North Dakota on October 5, 2026. California, Colorado, New York, and Texas have their own insurance AI rules or guidance instead (NAIC).

It can still reach you through your carriers. An insurer's written AI program "should address the use of AI Systems across the insurance life cycle," including "marketing," "underwriting," and "claim administration and payment." In an examination, regulators may ask for "Contracts with third-party AI System, model, or data vendors." A carrier that has to answer those requests may ask how your agency uses AI on its business.

Is health information on an application covered by HIPAA?

On health plan business, yes, and usually through the plan. HHS's own FAQ describes "insurance agents that are business associates of a health plan" (HHS). A business associate may pass protected health information to a subcontractor only after it "obtains satisfactory assurances" that the subcontractor "will appropriately safeguard the information," in practice a business associate agreement (45 CFR 164.502). OpenAI offers none for ChatGPT Business (OpenAI), and we cover which plans have one in private AI for pharmacists.

HIPAA's health plan definition leaves out excepted benefits (45 CFR 160.103), a list that includes "disability income insurance," "automobile liability insurance," and "Workers' compensation" (42 U.S.C. 300gg-91). HIPAA doesn't cover the medical details in a bodily injury claim or a disability income application once they're in the insurer's file. They are still nonpublic information under Model 668 and Part 500, which count health information whatever the line.

Can you paste an MVR or a claims report into ChatGPT?

Not as pulled. Two federal laws follow these reports. The Driver's Privacy Protection Act lets an insurer and its agents use motor vehicle records "in connection with claims investigation activities, antifraud activities, rating or underwriting," and a recipient may redisclose them "only for a use permitted" by the act. Anyone who rediscloses has to keep records for five years naming "each person or entity that receives information" and "the permitted purpose" (18 U.S.C. 2721). The act has no exception for sending a record to an AI company's servers.

If the report came from a consumer reporting agency, the Fair Credit Reporting Act adds its own limit. A person "shall not use or obtain a consumer report for any purpose unless" the purpose is a permitted one, such as "the underwriting of insurance involving the consumer" (15 U.S.C. 1681b). Asking an AI to explain a violation code doesn't need the driver's name, license number, or date of birth, so type the code and the dates and leave the rest out.

What does your carrier agreement say about AI?

The one we read doesn't mention AI, and its confidentiality terms reach it anyway. The Big I's review of an Auto-Owners agency contract (forms dated 7-22, reviewed August 2023) describes a "very broad confidentiality and data security addendum" whose confidential information includes "underwriting criteria," "underwriting rates and rules," "prospect lists," and the contract itself. The parties "shall not sell, retain, use or disclose the Data shared other than for the specific purpose of performing services specified in the Contract," and must report unauthorized access within 72 hours (Big I). A carrier's rate rules pasted into a chatbot to compare quotes fall under terms like these even when no client is named.

Utica National, which sells E&O coverage to agencies, asks for a written policy. Its June 2024 risk newsletter says "Agency management must develop a Generative AI Use Policy," and "Do not input this information into Generative AI tools that can potentially expose the information to the public" (Utica National).

What can an insurance agent paste into AI?

We read each rule's text on November 6, 2026 and applied it to the files an agent handles every day. "Consumer AI" means a personal ChatGPT, Claude, or Gemini account your agency hasn't vetted or signed a contract with.

How Morse keeps an agent's AI questions private

Every rule above turns on one fact. Once you press send, the client's information sits in readable form with an AI company, under that company's terms, and your agency answers for that company as its vendor. Morse is built so fewer readable copies exist. In E2EE, only a verified enclave can read your message, and Morse's servers relay ciphertext they can't read. In every mode, your chat history stays in your browser, not on our servers.

Whether a client's details may go into any tool, Morse included, is your agency's call under the rules and contracts above. Morse doesn't claim HIPAA compliance or any certification.

• E2EE, for claim and application questions, with the identifiers stripped. Your browser checks the secure enclave's hardware attestation, then encrypts the message so only that enclave can read it, and the reply is decrypted only in your browser. The enclaves are run by third-party providers, not by Morse, and the guarantee rests on the enclave hardware and attestation working as designed. E2EE runs open-weight models like GLM, Kimi, and DeepSeek without web search, so look up a statute or a filing in a separate chat.
• Private, for coverage questions, CE study, and drafting with no client in it. New chats start in Private, on open-weight models like DeepSeek, Kimi, and GLM, plus xAI's Grok. Morse has zero data retention agreements with its inference, search, and other providers, so they keep nothing. Where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't (what zero retention means).
• Anonymous, for marketing copy and general questions with GPT, Claude, or Gemini. Our proxy removes your account, email, and IP address and sends the message under Morse's own account. The provider sees the text without knowing it's you and may keep it for a limited time to check for abuse, under its own policy, so keep to the table's "Fine in a consumer AI" column there.

In every mode, Morse doesn't train AI on your content, sell your data, or show ads, so there's no training setting to turn off. Your history is stored in your browser, encrypted, and unlocking it takes two keys, one that never leaves your device and one our server gives you when you sign in. An attached file is sent with the message and never stored by Morse. Because we don't store chats, we usually have very little to give authorities (privacy policy). Clearing the browser, losing the device, or deleting your account deletes your chats, and Morse can't recover them, so anything that belongs in the client file goes into your agency management system. For a client asking about their own money, see private AI for financial questions.

Try a chat from the Morse home page without an account, or start free with no card. E2EE chats come with Pro, at $19 a month.