AI ethics guide · Texas
Can a Texas lawyer put client documents into an AI tool?
What Opinion 705 says about client confidentiality, and the other duties that come with using AI.
Not legal advice. This is a plain-English summary of Opinion 705, written by the team that makes ClientSafeAI, and it hasn't been reviewed by a lawyer. Read the source itself, and ask the Texas bar's ethics hotline if you're unsure how it applies to you.
| Document | Professional Ethics Committee Opinion No. 705 |
|---|---|
| Issued by | Professional Ethics Committee for the State Bar of Texas |
| Date | February 2025 |
| Status | Advisory ethics opinion |
The short answer
Yes, if you’re reasonably satisfied the tool won’t reveal confidential information to others. If you aren’t, Opinion 705 says that at a minimum you should not put confidential information into it without consulting the client and getting consent. The opinion singles out “self-learning” programs, which store user inputs and fold them into their datasets.
What it says about confidentiality
Opinion 705 builds on the Committee’s earlier cloud-computing opinion (Opinion 680), then goes further for generative AI because of how many of these programs learn from what users give them. It describes self-learning programs as ones that store user inputs and incorporate them into their existing datasets, where they could surface for someone else.
The test is practical. Before a Texas lawyer puts confidential client information into a generative AI program, the lawyer should be reasonably satisfied the program won’t reveal it to others. Without that, the minimum is not to input confidential information without client consultation and consent.
The opinion notes that ABA Formal Opinion 512 and Florida Opinion 24-1 both recommend informed consent before using generative AI with confidential information, and it follows that direction.
These concerns are especially relevant given the “self-learning” nature of many generative AI programs.
Texas Opinion 705, Confidentiality
The other duties it covers
| Competence | Understand hallucinations, gaps and errors in training data, and the confidentiality exposure before using a tool. |
|---|---|
| Supervision | Treat AI like a nonlawyer assistant under Rule 5.03. The lawyer is responsible for the work product whoever or whatever produced it. |
| Fees | Bill the actual time spent using and refining AI output, never time “saved”. The opinion adopts D.C. Opinion 388’s line that an hourly lawyer is never justified in charging for hours not actually expended. |
| Candor | Unverified AI output in a filing can implicate Rules 1.01, 3.01, 3.03 and 3.04. |
In practice
- For each AI tool, write down why you’re satisfied it won’t reveal client information, or don’t use it with client information.
- Treat any tool that stores inputs and learns from them as needing client consent.
- Check every citation an AI tool produces before it goes into a filing.
- Keep hourly bills to time actually worked.
How this compares with ABA Formal Opinion 512
The ABA's national opinion requires a client's informed consent before information relating to the representation goes into a self-learning AI tool, and says boilerplate consent in an engagement letter isn't enough. Read the Opinion 512 summary and firm checklist.
Keep client documents on a server that’s only yours
A Private AI Workspace runs on a GPU server ordered for one firm. Your documents, the AI model and the database stay on it, it never trains on your files, and answers cite the page they came from. It costs $499 a month, flat.
Sources
Spotted something out of date? Tell us and we'll fix it.
Not legal advice. This is a plain-English summary of Opinion 705, written by the team that makes ClientSafeAI, and it hasn't been reviewed by a lawyer. Read the source itself, and ask the Texas bar's ethics hotline if you're unsure how it applies to you.