Law Firm Chatbot and AI Intake Ethics: What You Can Automate and What the Bar Opinions Say

Not legal advice. Last reviewed 2026-10-14. We are not a law firm. State rules differ, and most of the opinions below are advisory. Confirm anything here with your state bar or ethics counsel before you rely on it.

Most attorneys ask about compliance first when AI comes up: confidentiality, unauthorized practice, data handling and privilege.

Those are fair questions, and bar regulators have now written a fair amount about them. None of the opinions summarized below bans AI intake. They attach conditions: understand the tool, supervise it, keep it from giving legal advice, tell people they are talking to software, and protect what prospective clients share.

This page has three parts: the ABA Model Rules that frame the question, a dated timeline of the bar guidance that bears on chatbots, voice agents and intake, and a practical breakdown of what AI can and can’t do in your intake, with a checklist at the end. Your state’s version of each rule is what actually applies.

The rules that frame the question

Five ABA Model Rules come up in almost every discussion of AI in legal intake. Every state has its own version, and the wording and numbering can differ.

Model Rule 1.1, Competence. Comment 8, added in 2012, says lawyers should keep up with “the benefits and risks associated with relevant technology.” For AI, that means understanding what a tool does with what people tell it.

Model Rule 1.6, Confidentiality. Lawyers must make reasonable efforts to prevent unauthorized disclosure of, or access to, client information. Any tool that touches client communications falls under this.

Model Rule 1.18, Duties to Prospective Clients. This is the intake rule. Someone who consults the firm about possibly hiring it is a prospective client, even if no engagement follows. The firm generally may not use or reveal what they shared. If the firm receives information that could be significantly harmful to that person, it can be barred from representing someone adverse in the same or a substantially related matter, unless an exception in the rule (such as written consent or screening) applies.

Model Rule 5.3, Nonlawyer Assistance. Lawyers with supervisory authority must make reasonable efforts to ensure nonlawyer assistance fits their professional obligations. That covers intake staff and vendors. ABA Opinion 512 applies it to how a firm’s lawyers and staff use AI tools, and the Oregon and New York City opinions below apply supervision duties to chatbots directly.

Model Rule 5.5, Unauthorized Practice of Law. A lawyer may not practice where not authorized, or assist another in doing so. Pennsylvania’s 2024 joint opinion and Alabama’s 2026 opinion both read this rule as reaching AI that gives legal advice without a lawyer’s involvement.

Two more rules show up in the chatbot opinions: Rule 7.1 (no false or misleading communications about the lawyer’s services, which is why several opinions want a chatbot to say it is a chatbot) and Rule 7.3 (solicitation).

Timeline: bar guidance on AI chatbots and intake (2010 to 2026)

Dates are the issue dates shown on each document or, where the document is undated, on the issuing body’s listing. Each line covers only what matters for chatbots, voice agents and intake; every opinion also covers other topics, such as billing and court filings. Most of these are advisory opinions, not binding rules, and an opinion from another state does not bind you, though it can show how regulators are thinking.

What the timeline adds up to. Read together, these documents point the same way. Label the bot as AI, not a lawyer. Keep it to facts and logistics, never advice. Collect only what you need, and get it into conflict checking fast. Have a lawyer able to see and correct what the bot says, and require lawyer approval before it takes any action on its own. Vet the vendor’s data handling in writing. Where they differ is in how prescriptive they are: Florida and the New York City Bar use “must” language about disclosure and supervision, while Oregon frames several of the same points as what would be “prudent.”

Voice agents. Most of these opinions say “chatbot,” but Oregon’s definition explicitly includes voice interactions, and the reasoning about disclosure, advice and supervision does not depend on whether the bot types or talks. Phone calls bring separate issues the bar opinions do not cover, such as call-recording consent and the FCC’s February 2024 ruling that AI-generated voices count as “artificial” voices under the TCPA, which restricts their use in outbound calls without prior consent. Treat those as a separate review with counsel.

New opinions keep appearing. If your state is not listed, check your state bar’s ethics opinions page for anything issued after the review date at the top of this page.

What AI can generally handle in law firm intake

Initial intake capture

Collecting basic information (who is involved, roughly what happened, when and where) is what a receptionist does every day. An AI intake system doing it works like a web intake form that responds right away and follows up. For a deeper split of tasks, see what to automate and what to keep human in law firm intake.

Generally treated as administrative because no legal analysis or judgment is applied. Florida’s Opinion 24-1 lists “obtaining factual information” as something nonlawyer intake staff may do. But mind Rule 1.18: collect only what you need to run a conflict check and route the inquiry.

FAQ responses

Answering standard questions about fees, practice areas, billing or what to bring to a consultation is usually administrative information, not legal advice.

Where to be careful: “Is my case worth pursuing?” calls for legal judgment. “How does a contingency fee work?” usually does not. Configure the AI to answer the second and send the first to an attorney, and have an attorney approve every scripted answer.

Appointment scheduling and follow-up

Booking consultations, sending confirmations and following up with prospective clients who haven’t responded.

Generally treated as administrative because scheduling involves no legal judgment. Follow-up messages still need to fit your state’s advertising and solicitation rules, and Oregon’s opinion is a reminder that automated outreach should not reach someone who is in no state to choose a lawyer.

CRM entry and routing

Capturing contact information, tagging by matter type and routing to the right attorney or paralegal: operational workflow, not legal work.

Review and referral requests

After a matter closes, automated review requests and referral prompts. Check your state’s rules on testimonials first.

What AI should not handle

Legal advice

The chatbot opinions in the timeline (Florida, Pennsylvania, New York City, Oregon, California and Alabama) draw the same line: a response that applies legal judgment to someone’s situation should come from a lawyer, not software. Pennsylvania and Alabama tie it to unauthorized practice under Rule 5.5, and the firm stays responsible for what its tools say.

“Based on what you’ve described, you likely have a strong case.” Most ethics counsel would treat that as legal advice. The AI should not say it.

“An attorney will review the details you’ve provided and follow up.” That is the kind of handoff the opinions describe. Route the judgment to a person.

In practice: configure responses to say they are not legal advice and that any assessment comes from an attorney, and test the configuration with the questions people actually ask.

The final conflict decision

The AI can prompt for what a conflict check needs: names of all parties, opposing counsel if known, related matters. It should not make the conflict determination.

Conflict analysis takes your full client database, judgment about related matters and professional responsibility. The AI gathers the inputs; a person makes the call. Oregon’s opinion adds the timing point: once a chatbot has enough about a prospective client, get them into the conflict system promptly.

Practical setup: the AI captures names and minimal facts first, then routes to designated staff for review before any substantive conversation continues.

Deciding what the conversation means legally

Whether an intake chat is privileged depends on your jurisdiction and the facts, so don’t assume either way, and don’t let the AI tell people. What is clear is that Rule 1.18 can protect what a prospective client shares even when no engagement follows. Treat intake data as confidential by default.

Prospective clients sometimes treat an intake chatbot like a consultation and share far more than you need. Florida and Oregon both warn that a chatbot that sounds like a lawyer can lead someone to believe a lawyer-client relationship exists. Configure the AI to explain its role: an automated assistant collecting basic information so the firm can follow up, not a lawyer giving advice.

Practical language (have your ethics counsel adapt it): “I’m an automated assistant for [Firm], not a lawyer or a member of the firm’s staff. I can’t give legal advice, and using this chat doesn’t make you a client. For now I only need your name, the names of others involved, and a one-sentence description of your matter. Please don’t share detailed facts yet; someone from the firm will follow up. Our privacy policy explains how we handle what you share.”

Acting on its own

Agent-style tools can send emails, texts or documents without being prompted each time. Oregon’s example is an agent that sends an engagement letter by itself, which the person could reasonably read as the firm taking the case. California’s 2026 guidance and Oregon’s opinion both point to the same control: no autonomous action that commits the firm without a lawyer’s approval.

Data handling

Under Rule 1.6 you need reasonable efforts to protect client information, and Opinion 512 and Oregon’s opinion both extend that to vetting AI vendors. For AI intake tools, that usually means:

Encryption in transit and at rest. Oregon’s opinion lists “ensuring that encryption is always used” among the steps that may be needed. Get it confirmed in writing.

Written terms on data use and retention. If client information passes through a third-party vendor, get written terms on how it can use the data, how long it keeps it, who can access it, and how you’ll learn of a breach. Oregon also tells lawyers to look past the vendor to the other software its product is built on.

No training on client data. Some AI services use conversation data to improve their models, and policies vary by product and plan. Oregon names keeping client information out of chatbot training as a likely safeguard. Confirm in writing that intake conversations aren’t used for training.

Transcript review. Oregon’s opinion assumes the lawyer can monitor what the chatbot says and change its responses. If a vendor can’t give you transcripts and a way to adjust answers, that is a gap to raise with counsel.

Setup checklist

Before deploying AI in legal intake:

• Ethics counsel has reviewed your state’s rules and any AI guidance from your state bar
• The bot says, at the start, that it is an AI program and not a lawyer or firm employee (Florida’s Opinion 24-1 uses “must”; New York City and Oregon point the same way)
• Responses state they are not legal advice and that using the bot does not make someone a client
• Disclaimers are conspicuous, plain and in the language the conversation is in (ABA 10-457)
• Intake collects only what conflict checks and routing need (Rule 1.18)
• Prospective clients go into the conflict system promptly, and conflict-relevant information gets human review before substantive conversation continues
• A responsible lawyer is notified of new chatbot conversations and can review and correct them (Oregon 2026-208)
• The bot cannot send engagement letters or take other committing actions without lawyer approval
• Escalation path for distressed, urgent or sensitive callers
• Written vendor terms on data use, retention, training and breach notice
• Encryption confirmed for data in transit and at rest
• Firm AI-use policy written and staff trained (ABA 512, Rule 5.3)
• Voice calls: recording-consent and AI-voice rules reviewed separately with counsel
• Attorney review of the AI configuration before launch (attorney review, not engineer review), and again whenever the vendor changes the product

The bottom line

The bar guidance so far allows AI in legal intake with conditions. It should do administrative work (information gathering, scheduling, routing, follow-up), identify itself as AI, and be supervised like any other nonlawyer assistance. The line is professional judgment, and a well-configured system hands everything past it to a person.

The risk comes from AI set up to do more than that: assessing cases, giving procedural advice, collecting detailed facts before a conflict check, acting on its own, or handling sensitive matters without an escalation path.

Related guides

• Law firm intake: what to automate, what to keep human, the task-by-task split this page’s rules apply to
• 5 signs your law firm’s intake process is costing you clients
• Why law firms lose clients after hours
• Website design for law firms, including where intake forms and chat sit on the page
• Clio Grow AI: what it does and what to ask before you turn it on
• AI intake for law firms: how it fits a legal practice

If you want a concrete plan to take to your ethics counsel, our AI Readiness Audit maps where AI could fit in your intake and where it should stop, before anything is built. The 2-week AI Readiness Audit starts at $2,500.