Can Your Firm's AI Chatbot Commit Unauthorized Practice of Law? What Attorneys Deploying Client-Facing AI Must Know
10/15/20265 min read
Can Your Firm's AI Chatbot Commit Unauthorized Practice of Law? What Attorneys Deploying Client-Facing AI Must Know
In early 2023, a startup announced that an AI "robot lawyer" would argue a traffic case in a real courtroom. Within weeks, after state bar officials reportedly raised the prospect of prosecution for unauthorized practice of law, the plan was pulled. It was an early signal of a question that has only grown more pressing as firms roll out client intake bots and website assistants: when does a client-facing AI cross the line into practicing law — and who is responsible when it does?
This isn't only a problem for legal-tech companies. Any law firm deploying a chatbot that interacts with the public is operating near two distinct risks, and both deserve attention before launch, not after a complaint.
Two different UPL problems
It helps to separate them clearly, because they're governed by different concerns:
The first is a client-facing AI that gives legal advice or holds itself out as a lawyer — the classic unauthorized-practice-of-law exposure, plus a misrepresentation problem. The second is an attorney over-delegating professional judgment to AI — effectively letting the tool, rather than the licensed lawyer, do the practicing. California is moving to address both.
What UPL actually is in California
California's rule is straightforward and strict. Under Business and Professions Code section 6125, no one may practice law in California unless they are an active licensee of the State Bar. Section 6126(a) makes it a crime to advertise or hold oneself out as entitled to practice law without being an active, licensed member. Practicing law — giving legal advice tailored to a person's situation, selecting or preparing legal documents for them, representing them — is reserved to licensed attorneys.
An AI system has no license. So a chatbot that dispenses tailored legal advice, tells a user which form to file, or presents itself as a lawyer is doing something no unlicensed person could lawfully do — and the liability lands on whoever deployed and controls it.
The cautionary tale: DoNotPay
The "world's first robot lawyer" became the template for how this goes wrong, and it drew fire from two directions at once.
On the state side, according to the FTC's complaint, the State Bar of California's Office of Chief Trial Counsel opened an unauthorized-practice-of-law investigation into the DoNotPay service around November 2021 and later sent a cease-and-desist notice. On the federal side, the FTC brought an action in September 2024 as part of its Operation AI Comply sweep, resolved by consent order: DoNotPay agreed to pay $193,000, to stop claiming its AI was an adequate substitute for a human lawyer, and to notify affected subscribers.
The important nuance: the FTC case was a deceptive-advertising action, not a UPL case — UPL is a state-level matter for bars and courts. But the two frameworks reinforce each other. The FTC established that marketing an AI as a "lawyer" it can't substantiate is a deceptive trade practice, while state bars police the unlicensed practice itself. A legal-AI product facing both at once is in a very hard place. The practical fallout was immediate across the industry: serious legal-AI vendors shifted their marketing from "AI lawyer" to "tool for legal professionals."
The firm-deployment angle: your intake bot
Most firms aren't building a robot lawyer. They're adding a website assistant or intake bot — and that's exactly where the line gets blurry. The safe zone is general legal information and logistics: explaining what practice areas the firm handles, scheduling a consultation, describing a process in general terms. The danger zone is tailored legal advice: analyzing a user's specific facts, telling them whether they have a claim, predicting an outcome, or recommending which document to file.
Disclosure matters too. Florida Bar Opinion 24-1 — persuasive guidance from another jurisdiction, not binding in California — requires that a client-facing chatbot identify itself as an AI and not as a lawyer, and advises obtaining informed consent before using third-party tools that would expose confidential information. California's own State Bar guidance recommends informing clients when AI is used in their representation. Across jurisdictions, the consensus direction is the same: a bot should never let a user believe they're talking to a lawyer, and it should not deliver what amounts to individualized legal advice without a licensed attorney in the loop.
The delegation line: you can't hand the practice of law to a machine
The second risk runs inward. California has advanced a bill that would make the principle explicit — providing that an attorney shall not delegate the practice of law to generative AI. As of recent reporting it was awaiting final action and is not yet enacted, so treat it as a pending development rather than current law. But it codifies something already implicit in existing duties: AI can assist a lawyer, but a licensed lawyer must exercise the professional judgment. Letting the tool make the legal decisions isn't efficiency; it's abdication — and it's the mirror image of the client-facing UPL problem.
A pre-launch checklist for client-facing legal AI
Before any firm bot goes live, pin these down:
Label the bot clearly and persistently as an AI, never as a lawyer or "legal counsel." Scope it to general information and logistics; block it from analyzing a user's specific facts, predicting outcomes, or recommending specific filings. Route anything that looks like a request for legal advice to a licensed attorney, with a clean handoff. Audit your marketing: drop "AI lawyer," "legal advice," or "as good as a lawyer" claims you can't substantiate. Keep a licensed attorney responsible for — and supervising — anything the tool does that approaches the practice of law. Treat disclaimers as necessary but not sufficient; a disclaimer won't cure a bot that actually gives advice. Monitor and log outputs so you can catch drift before a user relies on it.
FAQ
Can an AI chatbot commit unauthorized practice of law? The AI can't be licensed, so the exposure runs to whoever deploys and controls it. A bot that gives tailored legal advice or holds itself out as a lawyer can create UPL and misrepresentation liability for the firm or company behind it.
Was DoNotPay a UPL case? The FTC action was a deceptive-advertising case, not UPL. Separately, the California State Bar had opened a UPL investigation. The two tracks are distinct but mutually reinforcing.
Can my firm's intake chatbot answer legal questions? It can provide general legal information and handle logistics. It should not analyze a specific user's situation, predict outcomes, or recommend specific documents without a licensed attorney involved — and it must not present itself as a lawyer.
Does a disclaimer solve the problem? No. Disclosure that the user is talking to an AI is necessary, but a disclaimer won't save a bot that is actually dispensing individualized legal advice.
Can I just let AI handle routine legal work end to end? No. A licensed attorney must exercise the professional judgment. California has moved to state this expressly — and it already follows from existing competence and supervision duties.
Client-facing AI is one of the fastest ways to improve access and responsiveness — and one of the fastest ways to back into a UPL or misrepresentation problem if the guardrails aren't set first. If your firm is deploying an intake bot or client assistant, we help attorneys design it to stay on the right side of these lines before it ever talks to a client.