Texas AI Ethics Rules for Lawyers

The official instruments

A deliberate opinion from a deliberate state

Texas took its time. While early movers issued guidance through 2023 and 2024, the State Bar of Texas created a Taskforce on Responsible AI in the Law and let it work; Opinion 705, February 2025, is the Professional Ethics Committee’s response to that Taskforce’s request for a foundational statement. The opinion is explicit about being a high-level overview, which is best read as a floor with more specific guidance plausible later, and it compensates for brevity with unusual clarity: four duties, each tied to a specific Texas Disciplinary Rule.

Competence under Rule 1.01 means reasonably understanding the generative AI technology you use: what it does, where it fails, and what its terms do with your inputs. The opinion does not demand technical depth; it demands that the lawyer not be surprised by the tool’s basic behavior, hallucination included.

Confidentiality under Rule 1.05 is Texas’s distinctive framing, because Rule 1.05 protects both privileged and unprivileged client information, a broader field than many states’ Rule 1.6 analogues. Before that information is exposed to a generative AI tool, the lawyer evaluates the tool’s confidentiality posture and considers whether client consent is needed. For Texas firms the operational consequence is an approved-tools discipline that records each tool’s terms, plus consent language available for matters where exposure is real; the confidentiality guide supplies the tier framework.

Verification is where 705 is most pointed: independent verification of AI output before relying on it, with Mata v. Avianca cited by name. A formal opinion citing the sanctions caselaw directly converts “the machine sometimes lies” from cautionary tale into noticed duty. Texas practice adds a second layer to the same point: several Texas federal judges were among the earliest to adopt AI certification standing orders, so the per-filing forum check described in the court orders guide is not theoretical here.

Billing closes the square: efficiency gains belong to the client, and a lawyer may not bill for hours not worked. Combined with the national pattern (learning time as overhead, costs only by agreement), Texas hourly practice keeps speed gains only by changing fee structures, the move Virginia’s LEO 1901 later blessed explicitly; see the billing guide.

What Opinion 705 leaves to the rules themselves

The opinion’s restraint is informative. It does not mandate client disclosure categorically, does not regulate specific products, and does not write a supervision program, leaving Rules 5.01 and 5.03 to operate on their own terms for firm measures. The composite Texas posture is therefore: the four named duties from 705, plus standard supervision architecture (written policy, training, verification workflow) as the means of evidencing them. A Texas firm that adopts the policy template, sets its consent posture to clear Rule 1.05’s breadth, and enforces database verification on every filing has implemented everything the opinion asks and is positioned for whatever the Taskforce produces next.

The Texas checklist

Record each tool’s terms before use and keep the register current. Treat Rule 1.05’s broad definition seriously: unprivileged client information counts, so anonymization and tier discipline matter even for “harmless” inputs. Verify every citation in a traditional database before filing, and check the assigned judge’s standing orders each time. Bill actual time; restructure fees if you want the efficiency. And watch the tracker for Taskforce follow-on work, which is precisely the kind of development the changelog and alerts exist to catch.