Your Malpractice Carrier Is Now Asking About AI: A Michigan Lawyer’s Readiness Guide (2026)
If your firm’s professional liability renewal came up this year, you probably noticed something new: a page of questions about artificial intelligence. Do you use generative AI in your practice? Which tools? Do you have a written policy? Who checks the output? A few years ago those questions did not exist. Now they sit inside the underwriting file, and how you answer them can move your premium, shape your coverage, and in a bad case give a carrier a reason to fight a claim later. This guide explains what changed in the lawyers professional liability market, what the questionnaires are really testing, and the concrete steps a Michigan lawyer can take to answer them from a position of strength rather than improvisation.
Why the Questions Appeared
The trigger was the wave of hallucinated-citation cases. When lawyers filed briefs built on cases a chatbot invented, and courts responded with fee awards and referrals to discipline, insurers saw a new and measurable source of loss. The response has been quick. Carriers writing lawyers professional liability coverage have begun adding AI-specific questions to renewal applications, and the questions are no longer curiosities. They are underwriting tools, used to sort firms that manage the risk from firms that do not.
Insurance analysts describe the shift bluntly. Broker commentary on the 2026 professional liability market treats the period from early 2025 to early 2026 as a structural break, the point at which silence about AI disappeared from renewal paperwork and was replaced by affirmative questions, conditioned coverage, and in some placements outright exclusions. Whatever your personal view of generative AI, the market has already decided it is a rated exposure. The practical question for a Michigan practice is no longer whether to engage with these questions, but how to answer them without creating a problem you will regret.
What the Questionnaire Is Actually Testing
Read past the individual questions and a single theme runs through every carrier’s form: governance. The insurer is not really asking whether you use AI. It assumes you do, or soon will. It is asking whether your use is supervised, documented, and verifiable, or whether it is ad hoc and invisible. The recurring items map to a short list of controls.
- Which generative AI tools the firm permits, and whether their use is limited to approved tools.
- Whether the firm maintains a written AI use policy.
- Whether attorneys and staff receive AI-specific training.
- How AI-assisted work product is reviewed and verified before it leaves the office.
- Whether client-confidential information is protected when it is entered into a tool.
- Whether the firm has a plan for responding to an AI-related error.
Notice what those items are. They are not a technology audit. They are the ethical duties a Michigan lawyer already owes, translated into checkboxes an underwriter can score. A firm that has taken its professional obligations seriously will find most answers already in hand. A firm that has not will be answering to two audiences at once, the carrier today and, if something goes wrong, a court later.
The Ethical Backbone Behind Every Answer
Michigan has not adopted a separate rule for artificial intelligence, and it does not need one. The State Bar of Michigan has been clear that the existing rules already govern. In its AI FAQs for Attorneys, first published November 18, 2024, and in the June 11, 2025 report of its AI Task Force, Transforming the Legal Profession in the Age of Artificial Intelligence, the State Bar treats MRPC 1.1 competence as the operative standard for AI use, while cautioning that the FAQs are neither legal advice nor a formal ethics opinion. The message is that competence now includes a working understanding of the tools you rely on, including their capacity to fabricate.
The national framework points the same direction. In ABA Formal Opinion 512, issued July 29, 2024, the American Bar Association walked through the Model Rules that generative AI implicates: competence, confidentiality, communication, candor to the tribunal, reasonable fees, and supervision. On supervision, the opinion is direct. Managerial lawyers must establish clear policies about the firm’s permissible use of these tools, and supervisory lawyers must make reasonable efforts to ensure that lawyers and nonlawyers comply with their professional obligations when they use them. That is MRPC 5.1 and 5.3 applied to a new kind of assistant. The duty to supervise a nonlawyer does not disappear because the nonlawyer is software.
Put those pieces together and the malpractice questionnaire stops looking like an intrusion. It is a mirror of the rules. When a carrier asks whether you verify AI output, it is asking the same question MRPC 3.3 candor and MRPC 1.1 competence already answer. When it asks whether you have a written policy, it is asking whether you have done what Opinion 512 says managerial lawyers should do.
How the Two Regimes Line Up
The table below shows how the common questionnaire items map onto the professional duties a Michigan lawyer already carries. The overlap is the reason a single, well-built compliance effort serves both purposes.
| Carrier questionnaire item | The duty it mirrors | What satisfies both |
|---|---|---|
| Which AI tools does the firm use? | MRPC 1.1 competence; vendor diligence under MRPC 5.3 | An approved-tools list with a short diligence note on each vendor. |
| Is there a written AI policy? | Managerial duty under MRPC 5.1 and 5.3; ABA Op. 512 | A firm policy stating permitted uses, prohibited uses, and required verification. |
| Do staff receive AI training? | MRPC 1.1 competence; supervision under MRPC 5.1 and 5.3 | A short annual training with a sign-off you keep on file. |
| How is AI output verified? | MRPC 3.3 candor; MRPC 1.1 competence | A rule that a lawyer independently confirms every citation and factual claim. |
| How is client data protected? | MRPC 1.6 confidentiality | No confidential data in consumer tools; use accounts that do not train on your inputs. |
A Practical Readiness Checklist
You do not need a technology budget or a compliance department to answer these questions well. You need a handful of documents that reflect how a careful firm already works. The following steps take a small Michigan practice from improvised to defensible.
Write a one-page AI policy. It does not have to be elaborate. State which tools are approved, that confidential client information stays out of any tool that trains on inputs, that a lawyer verifies every AI-assisted citation and factual assertion before it goes out, and who in the firm owns the policy. A page a new associate can absorb in five minutes beats a manual no one reads.
Keep an approved-tools list with a diligence note. For each tool, record what it is used for and one or two lines confirming you checked how it handles data. This is the vendor-diligence record MRPC 5.3 contemplates and the carrier expects.
Train once a year and keep the sign-off. A brief session covering hallucination risk, confidentiality, and the verification rule, with a signature or short quiz retained, converts an assertion on the questionnaire into a record you can produce.
Make verification a documented habit. The single most important control is also the simplest: a human lawyer confirms the authorities and facts in any AI-assisted work. Note the check in the file. When a carrier or a court asks how you guard against fabricated citations, you point to a practice, not a promise.
Answer the questionnaire accurately. This matters as much as any control. An application is part of the insurance contract. Overstating your governance to secure a better rate can hand the carrier a rescission or coverage argument at the worst possible moment. If you do not yet have a policy, say so, then build one before the next renewal. Truthful answers backed by real documents are the goal.
Where This Is Heading
The questionnaire is the leading edge of a broader repricing. Expect the questions to become more specific, expect proof of governance to move from optional to expected, and expect some placements to condition coverage on having a policy at all. For firms that treat AI seriously, none of this is a threat. The same discipline that reassures an underwriter, supervised use, documented verification, and protected client data, is the discipline that keeps you out of the sanctions reports in the first place. The lawyers who struggle will be the ones who used the tools quietly and wrote nothing down. The readiness work is modest, and the time to do it is before the renewal notice arrives, not after a claim.
Frequently Asked Questions
Does my Michigan firm have to use AI to answer these questions?
No. If your firm does not use generative AI, you can say so, and many carriers offer a path for that answer. But be honest about tools that may already be in use through research platforms or drafting software, and revisit the answer each renewal, because informal use has a way of appearing without a decision.
Is a written AI policy actually required in Michigan?
No rule uses those exact words. But the State Bar of Michigan treats MRPC 1.1 competence as governing AI use, and ABA Formal Opinion 512 tells managerial lawyers to set clear policies on permissible use under the supervision rules. A short written policy is the practical way to meet those duties, and it is what the carrier is asking to see.
Can I lose coverage over how I answer?
The application becomes part of the insurance relationship, so accuracy matters. Overstating your governance can give a carrier grounds to contest coverage or rescind a policy after a claim. Understating current use is safer than overstating controls you have not built. The right move is a truthful answer supported by real documents.
What is the single most important control to have in place?
Independent human verification of AI output. A lawyer confirming every citation and factual claim before filing addresses the exact risk that produced the sanctions cases, satisfies MRPC 3.3 candor and MRPC 1.1 competence, and is the control carriers care about most.
How do I keep client information confidential when using these tools?
Keep confidential client information out of any consumer tool that trains on what you enter, use accounts or enterprise settings that contractually exclude your inputs from training, and document that choice. That practice tracks MRPC 1.6 confidentiality and answers the data-security line on the questionnaire.
Where can I find Michigan-specific guidance to build from?
Start with the State Bar of Michigan AI FAQs for Attorneys and the June 2025 AI Task Force report, and read them alongside ABA Formal Opinion 512. Together they give a small firm enough to draft a defensible policy without hiring a consultant, though none of them substitutes for judgment on your own facts.
Choose a firm that takes its work, and its technology, seriously.
Attorney Manny Chahal handles Michigan injury and no-fault claims with the same verification discipline this article describes: real review, real documentation, no shortcuts. Free statewide consultation. No fee unless we recover. Call 1-844-MCHAHAL.
Call 1-844-624-2425