Guides · insurance renewal

What your malpractice carrier asks about AI

The AI-hallucination sanctions made the news, and professional-liability carriers noticed. Renewal applications are starting to ask how a firm uses AI. Here are the kinds of questions to expect, and the difference between answering them with an assurance and answering them with a record.

Why carriers are asking now

A carrier prices risk. A firm that lets anyone paste client facts into any chatbot, with no policy and no review, is a different risk than one that screens inputs, verifies output, and keeps proof. Until recently a carrier had no way to tell those two firms apart. The questionnaire is how they are starting to.

What carriers said in public this fall

Two reports from the fall of 2026 sum up where carriers are. Insurers interviewed for a Daily Record report (August 2026) urged firms to adopt a firm-wide AI policy rather than leave it to each lawyer, to keep confidential client information out of AI tools, and to have lawyers check AI-generated work themselves. A South Carolina Lawyers Weekly report (October 12, 2026) found carriers watching closely but not yet broadly raising premiums or excluding AI claims. That leaves a window: the firms that can show their practices now are the ones in a position to answer when the questions get sharper.

The kinds of questions to expect

These vary by carrier, but they cluster around the same handful of concerns. For each, the weak answer is a checkbox, and the strong answer is something you can produce.

Does the firm have a written AI-use policy?
Strong answer: yes, and here it is, with dated staff acknowledgements.
How does the firm keep client data out of AI tools?
Strong answer: a confidentiality screen runs before anything is sent, and only approved tools are used for client work.
How is AI output reviewed before it is filed or sent?
Strong answer: a lawyer verifies citations and factual assertions and signs off on the record, matter by matter.
Which AI tools are approved, and who oversees their use?
Strong answer: a maintained approved-tools list and a named supervising lawyer.
Can the firm produce a record of its AI use if asked?
Strong answer: yes, a per-matter, tamper-evident record, exportable on request.

Assurance versus evidence

Every one of those strong answers has the same shape: not "we are careful," but "here is the proof we were careful on this matter, on this date, by this lawyer." That is the answer that holds up when a claim is actually filed, which is the moment the questionnaire was really about.

Answer with a record

Candor produces the artifact the questionnaire is asking for.

A written policy with acknowledgements, an approved-tools list, screening before the model, verification and attorney sign-off, and a per-matter export you can hand to a carrier at renewal. Built as your team works, not reconstructed the week the application is due.

This is general information, not legal or insurance advice, and carrier applications differ. Answer your own carrier's questions accurately and completely, and consult your broker or counsel on specifics. Candor is software, not a law firm and not an insurer.