AI Chatbots Giving Legal Advice: What Michigan Injury Victims Should Know (2026)
After a Michigan car accident, the temptation to ask ChatGPT, Claude, Gemini, or Perplexity for a quick legal answer is real. The answers come back confident and detailed. Some of them are also wrong, outdated, or amount to the unauthorized practice of law. This guide explains where Michigan draws the line under MCL 600.916, what the State Bar of Michigan has said about AI ethics, and the specific risks injury victims face when they rely on an AI chatbot instead of a licensed attorney.
What an AI Chatbot Actually Is (And Is Not)
A general-purpose AI chatbot is a large language model trained on text. It predicts the next word in a sequence based on what came before. It does not have an attorney–client relationship with the person asking, does not owe a duty of confidentiality under MRPC 1.6, is not bound by competence under MRPC 1.1, and cannot be sued for malpractice. When a Michigan injury victim types “Do I have a no-fault case?” into a chatbot, the system produces a plausible-sounding answer that may or may not match the current state of Michigan law.
This matters because Michigan no-fault is a moving target. The 2019 reform package alone amended dozens of sections, and the Michigan Supreme Court continues to reshape the field through opinions like Andary v USAA, 512 Mich 207 (2023), Sherman v Progressive Michigan Insurance Co, No. 167826 (Mich April 2026), and Kandil-Elsayed v F & E Oil, 512 Mich 95 (2023). A chatbot trained on older text often returns the pre-reform rule with full confidence.
Where Michigan Draws the Line: MCL 600.916
Michigan makes the unauthorized practice of law a misdemeanor under MCL 600.916. The statute prohibits anyone who is not an active member of the State Bar from engaging in the law business or holding themselves out as authorized to give legal advice. The question now in front of courts and bar regulators across the country is whether a chatbot that produces personalized “what should I do” answers for a specific injury fact pattern crosses that line.
The State Bar of Michigan AI Report
In June 2025 the State Bar of Michigan Board of Commissioners’ AI Workgroup released Transforming the Legal Profession in the Age of AI. The report reaffirms the existing rules — MRPC 1.1 (competence), MRPC 1.6 (confidentiality), MRPC 5.3 (responsibilities regarding nonlawyer assistance), MRPC 5.5 (unauthorized practice) — and applies them to generative AI. Two themes are important for injury victims:
- Verification is non-negotiable for lawyers. Attorneys are required to verify every AI-generated citation, holding, and statutory reference before relying on it. Mata v Avianca, 678 F Supp 3d 443 (SDNY 2023), set the national tone; subsequent Michigan-area and Eastern District sanctions cases have continued it. If a licensed lawyer must check every word, an unrepresented claimant has no realistic way to do the same.
- Confidentiality is at risk. Free public chatbots typically reserve the right to use user inputs to train future models. Personal medical history, accident facts, settlement-offer details, and insurer correspondence pasted into a public chatbot may not stay private — and may be discoverable if the case ends up in litigation.
Common Wrong Answers Injury Victims Get From Chatbots
| The Chatbot Often Says | The Michigan Reality (2026) |
|---|---|
| “You have 2 years to file a Michigan PIP claim.” | The one-year-back rule under MCL 500.3145 limits recoverable losses; the 3-year tort SOL under MCL 600.5805 is a separate clock. The interplay is fact-specific. |
| “Open and obvious is a complete defense in slip and fall.” | Kandil-Elsayed abrogated Lugo v Ameritech in 2023; open and obvious is now a comparative-fault consideration under MCL 600.2959, not an automatic bar. |
| “Motorcyclists cannot get PIP in Michigan.” | MCL 500.3114(5) provides a priority order for motorcycle PIP; Mary Free Bed Rehabilitation Hospital v Esurance Insurance Co, ___ Mich App ___ (2026), clarified exhaustion rules. |
| “You don’t need a notice for a road-defect case.” | MCL 691.1404 imposes a 120-day notice deadline; missing it has dismissed many cases. |
| “Insurers must always pay attendant care.” | MCL 500.3157(10) imposes a fee schedule and the 56-hour family-provided care cap for many policies issued after July 1, 2021. |
Each of these errors is the kind of thing a careful claimant would never miss with a Michigan-licensed lawyer reviewing the file. None of them are obvious to a non-lawyer reading a fluent chatbot answer.
The Confidentiality Trap
Anything a claimant types into a public chatbot becomes a record. In litigation, the insurer’s lawyers can serve a request for production asking for “all communications and inputs to AI tools regarding your injury or claim.” Some courts have already required production. A defense lawyer reading a claimant’s own chatbot transcripts can compare them to deposition testimony, social-media posts, and medical records. Inconsistencies become impeachment.
The safer choice when researching an injury claim is to use the chatbot for general background, never for personal facts, and to keep the specific medical, financial, and accident details to communications with a licensed lawyer where attorney–client privilege (recognized under MRE 501 and MCL 767.5a(2)) applies.
When AI Can Help (And When It Cannot)
The point is not that AI is useless. It is that the right job for the chatbot is narrow.
- Useful: understanding what a term like “PIP” or “tort threshold” means in general; getting a starter list of questions to ask a lawyer; translating insurer correspondence into plain English.
- Risky: asking for a “settlement value” estimate; asking whether to sign a release; asking whether a claim is worth filing.
- Off-limits: drafting an actual demand letter, a release, or a court filing; pasting the entire insurer file in for an opinion; relying on a chatbot citation in a settlement negotiation.
Frequently Asked Questions
Can ChatGPT be my lawyer in Michigan?
No. ChatGPT is not licensed to practice law in Michigan or any other state. Under MCL 600.916, the practice of law in Michigan is restricted to active members of the State Bar. A chatbot can give you general legal information, but it cannot give you the kind of strategic, ethical, privileged representation that an attorney–client relationship provides.
Is what I type into a chatbot privileged?
No. Attorney–client privilege in Michigan (recognized under MRE 501 and MCL 767.5a(2)) applies to confidential communications with a licensed lawyer for purposes of legal advice. A chatbot is not a lawyer, so the inputs are not privileged. Worse, they may be stored, used for model training, or produced in discovery.
If a chatbot gives me wrong legal advice and I get hurt, can I sue it?
That is the open question in cases like Nippon Life v OpenAI. Product-liability theories are being tested. As of this writing, no Michigan court has held a chatbot vendor liable for an individual claimant’s bad legal outcome.
Should I show my lawyer my chatbot conversations?
Yes, if you have them. They can become discoverable, and your lawyer needs to know what the other side may eventually see. The earlier this comes out, the easier it is to address.
Are AI-drafted demand letters a good idea?
Generally not for unrepresented claimants. They tend to be over-confident, miss Michigan-specific elements like the no-fault interest provision under MCL 500.3142, and signal to insurers that the claimant is self-representing. See our prior guide on AI-Drafted Demand Letters for the practical risks.
What is the safest way to use AI after a Michigan accident?
Use AI for general background only. Keep specific personal facts, accident details, and medical history out of public chatbots. Get a licensed Michigan attorney involved early. For most claimants there is no fee for a consultation, and many no-fault cases are taken on contingency.
Injured in Michigan? Don’t face this alone.
Free, confidential review with Attorney Manny Chahal. No fee unless we recover.
Call 1-844-624-2425