AI Chatbots Posing as Lawyers: Michigan UPL Risk in 2026
Generative AI chatbots that draft demand letters, predict case outcomes, and tell consumers what to plead are everywhere on social media in 2026. When a chatbot crosses the line from general information to specific legal advice for a particular Michigan matter, the operator faces real unauthorized-practice-of-law (UPL) exposure under MCL 600.916, and the consumer relying on its output has no malpractice carrier and no privilege protection.
The Viral Hook Driving 2026 Attention
Social media is saturated with short videos of consumers running their car-accident facts through general-purpose chatbots and posting the output as a script for what to tell an adjuster or even what to file. A second wave of content shows lawyers reacting to that output, often pointing out it cites cases that do not exist or misstates Michigan law.
The Federal Trade Commission squarely addressed this issue with DoNotPay, which marketed itself as the “world’s first robot lawyer.” The FTC’s consent order required DoNotPay to pay $193,000 in monetary relief, notify affected subscribers, and stop advertising that its service performs like a real lawyer without adequate supporting evidence. The lesson for Michigan consumers is simple: an AI tool is not a licensed attorney, and a chatbot conversation is not legal representation.
Michigan’s Unauthorized-Practice Framework
Michigan law prohibits a person from practicing law, engaging in the law business, or holding themselves out as authorized to practice law unless regularly licensed and authorized to practice in Michigan. Under MCL 600.916, a person who violates this provision is guilty of contempt of the Michigan Supreme Court and of the circuit court of the county where the violation occurred.
The State Bar of Michigan’s UPL guidance draws the practical line: a nonlawyer may provide general forms and general information, but may not tell a person how to handle a case, predict a legal result, choose a legal form for a person’s situation, draft documents requiring legal judgment, tell someone what to say in court, or advise whether to accept a settlement.
A general legal-information tool that says “here is what a Michigan no-fault PIP claim looks like” is informational. A chatbot that takes a user’s facts, applies Michigan no-fault law, and tells the user how to file or what to demand is much closer to applying law to facts for a specific person. The line is not bright, and courts have not yet drawn it definitively for chatbots, but the analytic framework is well-established.
The controlling Michigan rule: MCL 600.916 makes the unauthorized practice of law contempt of court and authorizes injunctive relief. The prohibition is not limited to natural persons. Companies that hold themselves out as providing the equivalent of legal services in Michigan, including through software, face UPL exposure.
What ABA Formal Opinion 512 Actually Says
The American Bar Association issued Formal Opinion 512 on July 29, 2024, on the use of generative AI by lawyers. The opinion does not authorize chatbots to practice law. It governs how lawyers use AI tools within an existing attorney-client relationship. Key points: a lawyer remains responsible for competent representation under Model Rule 1.1, must maintain confidentiality under Rule 1.6 when feeding client information to a tool, must communicate with the client about AI use under Rule 1.4, must charge reasonable fees under Rule 1.5, and must supervise the tool under Rules 5.1 and 5.3.
Opinion 512 implies a corollary sometimes lost in social media coverage: if an AI tool is operated by a non-lawyer entity directly to consumers, none of those protections apply, because there is no licensed attorney inside the relationship to be responsible. That is precisely the gap UPL law is designed to fill.
The Hallucinated-Citation Problem Under Mata v Avianca
The most-cited example of AI-generated legal work going wrong remains Mata v Avianca, Inc, 678 F Supp 3d 443 (SDNY 2023), where Judge P. Kevin Castel sanctioned counsel under Rule 11 after a ChatGPT-generated brief cited multiple fabricated cases. The decision is not binding outside the Southern District of New York, but it is now routinely referenced in standing orders nationally, and many courts have since adopted orders requiring counsel to certify and verify AI-generated content before filing.
For consumers, the same hallucination risk exists. A chatbot can fabricate a realistic-looking case citation that sounds confident and authoritative. A claimant who quotes that fake case to an adjuster, or pleads it in court, undermines a legitimate claim and may face sanctions even as a self-represented party.
Practical Line for Michigan Claimants
A chatbot can do useful things without crossing the line: explain general legal frameworks, summarize what statutes say, organize a timeline, and surface questions to ask a licensed lawyer. It cannot competently apply Michigan no-fault, premises liability, or product liability law to a specific accident and produce a strategy that holds up in front of a real adjuster or judge.
Consumers who rely on chatbot output for case-strategy decisions have no malpractice carrier behind them, no attorney-client privilege protecting their input from discovery, and no professional supervision verifying that the cases cited actually exist. Those three protections are why Michigan licenses lawyers in the first place.
Risks for AI-Product Operators Marketing Into Michigan
Operators of AI legal-help products that market to or are accessible by Michigan consumers face overlapping exposure. The State Bar can pursue UPL enforcement under MCL 600.916. The Michigan Attorney General may pursue deceptive conduct under the Michigan Consumer Protection Act, MCL 445.901 et seq. And the FTC can act under Section 5 if marketing claims overstate what the product does, as the DoNotPay matter demonstrates.
Three product design choices materially reduce that exposure: a prominent disclaimer that the tool provides general information, not legal advice, and does not create an attorney-client relationship; a hard refusal pattern when a user asks for strategy on a specific named matter; and a built-in handoff to a licensed Michigan attorney for any output that approaches the application of law to particular facts. A disclaimer helps, but it is not dispositive — courts and regulators look at what the product actually does, not just what its footer says.
RELATED KNOWLEDGE BASE ARTICLES
Michigan Motorcycle Accidents: No-Fault PIP Priority (2026) · Serious Impairment Threshold After McCormick · PIP Coordination of Benefits: MCL 500.3109a
Frequently Asked Questions
Can I use ChatGPT to write my demand letter?
You can use it to draft a starting outline or summarize what a demand letter usually contains. Sending an AI-drafted letter to an insurer without a licensed Michigan attorney reviewing the case-specific facts, the statutes that apply, and the cases it relies on is risky. Adjusters increasingly recognize chatbot-style language, and a letter that cites a fabricated case undermines the entire claim.
Is a chatbot answer privileged?
No. Attorney-client privilege protects confidential communications with a licensed attorney made to obtain legal advice. A chatbot is not a licensed attorney, the conversation is not confidential in the legal sense, and the input is often retained by the service provider. Anything a consumer types into a chatbot may be discoverable later.
What if the chatbot says it is not legal advice?
A boilerplate disclaimer is one factor a court weighs but is not by itself dispositive. Michigan UPL analysis looks at what the tool actually does, not what its disclaimer says. A chatbot that pairs a “not legal advice” disclaimer with output applying Michigan no-fault law to a specific accident still carries UPL risk for the operator.
Can my lawyer use AI on my case?
Yes, within the framework ABA Formal Opinion 512 sets out. Your lawyer must remain responsible for the work product, verify any citations the tool produces, protect your confidential information, and communicate with you about how the tool is being used. Reasonable AI-assisted work is increasingly common and is not, by itself, malpractice.
What should I do if I already sent a chatbot-drafted letter?
Stop adding to it. Save the chatbot conversation so you know what facts the tool was given. Consult a licensed Michigan attorney before the next step. A bad first letter is usually salvageable if a real lawyer takes over the file before the insurer formalizes a denial.
Have a question about this area of Michigan law?
Free, confidential review with Attorney Manny Chahal. Consultations available in English, Hindi (हिंदी), and Punjabi (ਪੰਜਾਬੀ).
Call 1-844-MCHAHAL for a free case evaluation