What question the opinion answers
The question was simple: may Oregon lawyers use chatbots to communicate with clients or potential clients on their websites or elsewhere? The opinion covers both rule-based chatbots that follow fixed scripts and AI-powered ones built on large language models. It also addresses AI agents, which can act on their own toward a goal, and says they carry more risk because they can take steps no one prompted.
The opinion is explicit that lawyers are responsible for chatbots they use or make available to the public, whether on a website, a phone, by text or otherwise.
The rules it applies
| Rule | What the opinion says about chatbots |
|---|---|
| RPC 1.1 Competence | You may use a chatbot only after taking reasonable steps to understand how it works, its benefits and its risks. That evaluation happens at purchase and should be revisited as the product changes. |
| RPC 1.6 and 1.18 Confidentiality and prospective clients | Know how the tool stores information. Confirm data is encrypted, kept out of AI training, separated between clients and protected from other systems. Information from prospective clients is protected even if they never hire you. |
| RPC 1.18(c) Conflicts | Disqualifying information received through a chatbot can bar you from representing someone adverse. Enter prospective clients into your conflict system promptly. |
| RPC 5.3 Supervising nonlawyers | Treat the chatbot vendor like any outside provider: check references, credentials and security practices, and use confidentiality agreements. Due diligence extends to the other software the vendor's product is built on. |
| RPC 7.1 Misleading communications | A chatbot that doesn't identify itself as a chatbot could be misleading. It must not overstate your expertise, experience or results, and it must not give legal advice. |
| RPC 7.3(a) Solicitation | A chatbot may not notice that someone's physical, emotional or mental state prevents them from making a reasonable choice about hiring a lawyer. Lawyers should review those conversations and interview each prospective client before going forward. |
| In re Weidner (1990) | In Oregon, an attorney-client relationship can form based on the client's reasonable belief. A chatbot that sounds like a lawyer could create one without you knowing. |
The four practical risks
1. Accidentally forming an attorney-client relationship
The opinion recommends clear disclaimers on the website and in the chat. It also says it would be prudent for a responsible lawyer to be notified automatically after each chatbot conversation and to review it promptly to decide whether a relationship formed, whether other duties were triggered and whether to take the matter or tell the person you won't. A footnote notes that trained staff can handle that review.
2. Saying something false, misleading or advisory
The chatbot should be carefully programmed and the lawyer should review its conversations so misstatements can be corrected quickly. Answering "how much is my case worth?" or "do I have a case?" is the kind of response to block.
3. Leaking confidential information
A chatbot company may be built on other companies' AI models and connected to other services. Information can reach those third parties beyond what the chatbot company's own policy says, so due diligence has to cover the whole chain.
4. AI agents acting on their own
The opinion's example is an agent that sends an engagement letter because it decided that would help sign a client. Its guidance: subject autonomous actions, such as sending emails, letters or texts, to lawyer approval. Lawyers remain responsible for the accuracy of everything a chatbot produces.
A checklist before you turn on a chatbot
- The chatbot says it's an AI in its first message and confirms it if asked.
- Your website and the chat carry a clear disclaimer that no attorney-client relationship forms until a signed agreement.
- It collects names and opposing parties first, and leads go into your conflict system promptly.
- It declines to answer questions about case merits, value, deadlines or strategy and offers a consultation instead.
- You get every transcript, and a named lawyer or trained staff member reviews each conversation promptly.
- Nothing that could form a relationship, such as an engagement letter or fee agreement, goes out without lawyer approval.
- You have the vendor's list of underlying providers, a written data processing agreement and confirmation your data isn't used for training.
- Someone at the firm can explain how the tool works, and you revisit that review when the product changes.
- You have a plan for conversations where the person seems distressed, injured or not in a state to decide, so a person follows up instead of the bot pushing for a booking.
Questions to ask a chatbot vendor
- Which AI models and other companies process our visitors' messages?
- Is our data ever used to train your models or anyone else's?
- Where is data stored, is it encrypted and how is it separated from other customers?
- Can we see and change exactly what the chatbot says?
- Will it ever send emails, texts or documents on its own?
- Do we get every transcript, and how quickly?
- Will you sign a data processing agreement with breach notice?
- What happens to our data if we cancel?
How 2026-208 fits with Oregon's other AI opinion and ABA Opinion 512
Opinion 2026-208 is Oregon's second formal opinion on AI. The first, OSB Formal Opinion 2025-205 (February 2025), covers AI tools in law practice generally. The American Bar Association's Formal Opinion 512 (July 2024) covers generative AI tools nationally. Firms that use a chatbot for intake are working under all three.
| OSB 2025-205 | OSB 2026-208 | ABA 512 | |
|---|---|---|---|
| Covers | AI tools lawyers use in their own work | Chatbots and AI agents that talk to clients and prospective clients | Generative AI tools in law practice, nationally |
| Main concern | Competence, confidentiality, fees, supervision, candor to courts | Prospective clients, accidental relationships, misleading statements, vendor supervision | Competence, confidentiality, communication, candor, supervision, fees |
| Who it binds | Guidance for Oregon lawyers | Guidance for Oregon lawyers | Persuasive guidance; Oregon's own opinions come first |
Our blog post Oregon's two AI ethics opinions, explained walks through 2025-205 and 2026-208 side by side, with citations.
How Tactical Intake handles it
We built our AI intake chat and after-hours call answering around this opinion: AI disclosure first, conflicts before facts, no legal advice, no autonomous engagement letters, every transcript to the firm, a Data Processing Addendum and a 20-scenario test before launch that your supervising attorney signs off on.
Source: OSB Formal Opinion 2026-208, "Chatbots and AI Agent Communications", approved by the Board of Governors, February 2026. The opinion also refers to OSB Formal Opinion 2025-205 on AI use generally. This page is a summary, not legal advice. Your firm is responsible for its own compliance, and the OSB offers an ethics helpline for members.
Common questions
Can Oregon lawyers use AI chatbots on their websites?
Yes, with qualifications. OSB Formal Opinion 2026-208 says lawyers may use chatbots and AI agents if they understand the technology, put safeguards in place and meet their duties under the Rules of Professional Conduct.
Does a law firm chatbot have to say it's a bot?
The opinion notes that more than one state has found a chatbot that doesn't identify itself as a chatbot could be misleading under RPC 7.1. Disclosing it in the first message is the safe practice.
Can a chatbot create an attorney-client relationship?
It can. In Oregon the relationship depends on the client's reasonable belief, so a chatbot that sounds like a lawyer could create one. The opinion recommends disclaimers and prompt lawyer review of chatbot conversations.
When was OSB Formal Opinion 2026-208 issued?
The Oregon State Bar's Board of Governors approved Formal Opinion 2026-208, "Chatbots and AI Agent Communications," in February 2026.
Can a law firm chatbot send an engagement letter on its own?
The opinion uses that as its example of AI agent risk. Its guidance is to subject autonomous actions, such as sending emails, letters or texts, to lawyer approval.
How is Opinion 2026-208 different from Opinion 2025-205?
Opinion 2025-205 (February 2025) covers AI tools lawyers use in their own work. Opinion 2026-208 (February 2026) covers chatbots and AI agents that communicate with clients and prospective clients on the firm's behalf.
Which Oregon rules does Opinion 2026-208 discuss?
RPC 1.1 (competence), 1.6 and 1.18 (confidentiality and prospective clients), 5.3 (supervising nonlawyers and vendors), 7.1 (misleading communications) and 7.3(a) (solicitation of people who can't exercise reasonable judgment).
See it working on a real firm site
Chat with the assistant on our demo firm site, then book 15 minutes and we'll show you where the lead lands. Questions? Call (775) 762-6203.