Does a law firm's malpractice policy cover AI drafting errors?
For the client's claim, usually yes: a negligent draft is a negligent draft. The court's sanction is a different story, and the answer there is usually no. That gap is where firms get surprised.
Answered in short
5 things that decide this
- 01A lawyers' professional liability policy covers negligent acts, errors and omissions in legal services. A brief or contract clause drafted with AI help and filed under your firm's name is a legal service, so a client's claim over a drafting error usually sits inside the insuring clause.
- 02Most malpractice policies in force in 2026 are silent on AI. They neither confirm nor exclude it. Coverage lawyers call this gap 'silent AI', and it means the outcome rests on definitions and exclusions written for other risks.
- 03Most lawyers' policies exclude fines, penalties and sanctions. A court's sanction for a fabricated citation is therefore usually your firm's own cost, even when the client's related claim is covered.
- 04ABA Formal Opinion 512, issued July 29, 2024, makes the lawyer accountable for AI output under the existing duties of competence and supervision. Coverage follows the same logic. The policy pays for your negligence, and a supervision record proves the error was negligence rather than recklessness.
- 05Check your last renewal application for an AI question. A firm that answered 'no AI use' while associates drafted with a chatbot has an application problem, and that can cost coverage before any exclusion is read.
Three parts of the policy decide the answer
Your policy has three parts that decide coverage for an AI drafting error: the insuring clause, the exclusions and the application. The model that wrote the draft appears in none of them. That's the whole silent-AI problem.
Start with the insuring clause, which is on your side. Drafting a will, a lease or a motion is a professional service, and your carrier agreed to pay for negligence in professional services. A wrong clause produced by a model and released unread is exactly that. The tool matters no more than the word processor did.
Exclusions are where the surprises live. Nearly every lawyers' policy excludes fines, penalties and sanctions, so a court's order over a hallucinated citation lands on your firm. A carrier can argue the dishonest-acts or knowing-violation exclusion when a lawyer certified a filing without reading it. And some forms gained an AI-specific exclusion in the 2026 renewal cycle. If yours did, the question is closed.
Your application is the quiet third part. Carriers began asking firms about AI use and AI policies at renewal. If your answer doesn't match what happens at the desk, the carrier has a misrepresentation argument. That is a worse position than any exclusion.
What a firm does about it
Ask your carrier in writing how an AI-assisted drafting error would be treated under your form, and keep the answer with the policy. Then read the exclusions page yourself. Look for sanctions, fines, technology and automated. Those two steps turn silence into a position you can plan around.
Then make supervision provable. Supervision is what separates a covered negligent error from an uncovered reckless one. ABA Opinion 512 applies the competence and supervision duties to generative AI, and 15 US jurisdictions had issued formal AI ethics opinions by June 2026. Your carrier reads those the way a bar examiner does. Was there a review, and can you show it?
A supervision record is a build, not a memo. Every AI-assisted draft passes a required human check before it leaves your firm, and the system logs who checked it and when. LexPair, a legal platform Hashlogics built, uses AI-assisted logic to sort each case to the right practice area. Its admins keep full visibility into every lead. Point that same visibility at drafting, and your managing partner has the record a claims adjuster asks for on the first call.
- Draft generatedFrom templates and matter facts.
- Citations checkedEvery source opened, not skimmed.
- Lawyer signs offNamed reviewer, before release.
- LoggedWho, when, what changed.
- Claim arrivesNegligence, not recklessness.
The log doesn't make the error go away. It decides which exclusion the carrier reaches for.
Related questions
01Does malpractice insurance cover a sanction for a hallucinated citation?+
Usually not. Most lawyers' policies exclude fines, penalties and sanctions, so the court's order is your firm's own cost. The client's claim arising from the same filing may still be covered as negligent professional services. Read the sanctions exclusion on your own form.
02Will telling our carrier we use AI raise our premium?+
Answer the application truthfully, whatever it does to the premium. A misstatement about AI use gives the carrier a rescission argument that removes all coverage, and that costs more than any surcharge. A documented review process is what keeps an underwriter comfortable.
03Is a client's data leak through an AI tool a malpractice claim?+
Often it's a cyber claim, and many malpractice policies exclude data breaches or cap them at a small sublimit. In United States v. Heppner, in February 2026, a federal court found no privilege in chats with a consumer AI tool whose terms allowed training. Check that your cyber and malpractice policies meet with no gap.
Related
- What happens when a lawyer cites a hallucinated case? →The sanction side of the same error.
- What does ABA Formal Opinion 512 require? →The accountability baseline carriers read from.
- What is silent AI in an insurance policy? →The wording gap, defined.
- Legal document automation →Drafting from matter facts with a review gate built in.
