Insuring against "productive laziness": AI in law and what your policy will - and won't - cover
Pressed for time, lawyers now reach for AI to draft letters and briefs in minutes. The speed feels like a win - until the citations don't exist, the reasoning collapses, and the court notices.
AI is predictive text with confidence, not legal judgment. Studies have documented high rates of hallucinations on specific legal prompts, and courts are responding. Use it, but treat every output as unverified until you prove it.
Why AI tempts busy lawyers
AI tools can produce clean prose fast. That's the trap. What reads well can be half-true, outdated, or flat-out fictional.
A Stanford-led body of work has flagged hallucinations as a persistent issue with large language models. See an overview from Stanford HAI on hallucinations here: What are AI hallucinations?
Or as one entrepreneur put it: a rocket ship without controls becomes a bomb. In law, the blast radius includes sanctions, discipline, malpractice claims, and billing disputes.
What courts are doing about it
Courts have sanctioned and publicly criticized attorneys for filing AI-generated briefs with hallucinated caselaw. Recent decisions (including Noland v. Land of the Free and In re Martin) imposed penalties and, in some instances, referrals to state bars for potential discipline.
One judge said it plainly: "At this point, no lawyer should be using ChatGPT or any other generative AI product to perform research without verifying the results." Expect verification - not excuses.
Your insurance snapshot
- Sanctions: Commonly excluded under professional liability policies. In some places, they're uninsurable as a matter of public policy (e.g., willful misconduct issues linked to California Insurance Code § 533). See the statute: Cal. Ins. Code § 533.
- Disciplinary matters: Many policies provide a modest sublimit for responding to disciplinary inquiries (often $20,000-$25,000 for defense fees).
- Malpractice (negligence): If AI-driven research or drafting causes a blown motion, appeal, or case, most policies cover it as a "Wrongful Act" - unless there's a specific AI exclusion. Policies insure the service, not the tool.
- AI exclusions on the rise: Some insurers now exclude liability "arising out of" the use of generative AI. If it's in your policy, coverage for AI misuse may be off the table.
- Fee disputes and fraud: Claims for fee return/overbilling are typically excluded. Fraud exclusions are standard. Don't expect coverage if billing inflated time for AI-generated work.
- Claims-made and reported: Coverage depends on prompt reporting within the policy period. Late notice can kill coverage - even for covered claims.
Risk controls that actually work
- Verify every authority: Treat all AI-cited cases and statutes as suspect. Shepardize/KeyCite and read the full text before filing.
- Log your review: Keep prompts, outputs, and human edits in the file. If challenged, you can show real legal work happened.
- Follow court rules: If a judge or local rule requires disclosure or certification, comply. Don't surprise the court.
- Fix billing: Bill the actual time spent. Cap AI-assisted drafting time, ban "value billing" for AI outputs, and document who did what.
- Lock down tools: Use approved platforms with confidentiality controls, audit logs, and access restrictions. Know where your data goes.
- Tune your policy and training: Review your malpractice policy for AI exclusions, disciplinary sublimits, panel counsel, and consent-to-settle terms. Train your team on verification and billing standards.
If you need structured AI training and guardrails for your team, see curated options by role: Complete AI Training - Courses by Job.
What to do the moment AI work product goes sideways
- Stop the bleeding: Withdraw or correct the filing. Do it fast.
- Preserve evidence: Save prompts, outputs, drafts, and verification notes.
- Inform the client: Early, direct, and documented.
- Tender to your insurer: Report as a claim or potential claim immediately. Include dates, deadlines, and any court orders.
- Consult ethics counsel: Assess disclosure duties, conflicts, and next steps.
Bottom line
AI can help you move quicker, but it won't carry your ethical and professional duties. Courts expect human verification. Your policy likely covers negligence - not sanctions or fraud - and new AI exclusions can change the outcome.
Keep control of the rocket ship: verify, document, bill honestly, and report problems early.
Your membership also unlocks:
