A Texas lawyer who uses generative AI answers to two authorities. Rule 1.01 governs competence, and the Texas Professional Ethics Committee reads it to reach technological competence 2. Courts govern filings, and they have said that existing rules make attorneys gatekeepers for the accuracy of their submissions and that "'I used AI' will not be an excuse for an otherwise sanctionable offense" 3, 4. Different bodies enforce the two duties, but both lead to one practice: verification. Competence means you know how the tool fails. Accountability means you catch the failure, or correct it openly, before it costs you. Confidentiality appears here only where the committee ties it to technological competence 2.
Competence: why tool skill becomes mandatory
The rule, the comment, and the opinion
Comment 8 to Rule 1.01, in the rules effective March 7, 2025, says each lawyer "should strive to become and remain proficient and competent in the practice of law, including the benefits and risks associated with relevant technology" and that "a lawyer should engage in continuing study and education" 1. That language urges rather than commands, and alone it gives a thin basis for discipline.
The mandatory language sits in the rule. As quoted in Opinion 705, Rule 1.01(a) provides, with limited exceptions, that a lawyer "shall not accept or continue employment in a legal matter which the lawyer knows or should know is beyond the lawyer's competence," and the Terminology section defines competence as the "possession or the ability to timely acquire the legal knowledge, skill, and training reasonably necessary for the representation of the client" 2.
Weigh Opinion 705 for what it is. The committee issued it in February 2025, at the request of the State Bar's Taskforce on Responsible AI in the Law, as a "high-level overview," and calls it "a snapshot of potential ethical concerns at the moment and a restatement of certain ethical principles for lawyers to use as a guide" 2. It is committee guidance and carries neither the force of rule text nor that of a court holding. Treat it as persuasive evidence of how to read Rule 1.01, and keep the rule itself in front of you.
The argument from the definition
The definition does the work, and three of its features make tool-specific skill mandatory once a lawyer chooses the tool 2. First, the definition measures necessity against "the representation of the client," meaning the work as the lawyer performs it. A lawyer who drafts with a generative model has made it part of the representation, so the skill to check its output is necessary. Second, the definition credits the "ability to timely acquire" skill, so a lawyer need not have expertise in advance, but must have verification skill by the time the output matters. Third, Rule 1.01(a) asks what the lawyer "knows or should know" 2. The committee has warned that many models "hallucinate," producing "inaccurate or made-up answers that sound convincing," and calls Mata the most famous sanctions example 2. After that warning, a Texas lawyer will struggle to call the risk unforeseeable.
The ABA frames the same question
ABA Formal Opinion 512 asks, under the Model Rules, "What level of competency should lawyers acquire regarding a GAI tool?" and "What level of review of a GAI tool's process or output is necessary?" 5. It calls GAI tools "a rapidly moving target" 5, which echoes the Texas snapshot caveat 2. Both bodies place review of output inside the competence inquiry, which supports reading Rule 1.01 to require verification rather than mere awareness. In Texas the ABA view persuades and does not control.
Accountability: courts police the filing
Mata: the sanction followed the cover-up
Mata v. Avianca is a Southern District of New York decision applying Federal Rule of Civil Procedure 11, so it persuades in Texas and does not bind 2, 3. The respondents kept standing by the fake opinions after judicial orders questioned their existence, and the court found bad faith by the individual respondents "based upon acts of conscious avoidance and false and misleading statements to the Court" 3. Had they come clean earlier, "the record now would look quite different" 3.
Keep the remedy's two bases separate. The court ordered sanctions under Rule 11 or, alternatively, its inherent authority, including a $5,000 penalty 3. Separately, it held the firm jointly responsible because Rule 11(c)(1) requires that result "[a]bsent exceptional circumstances," and it found none 3. The federal rule creates the firm's liability there.
The Fifth Circuit: no special rule, same duty
On November 22, 2023, the Fifth Circuit proposed requiring certification that no generative AI was used or that "all generated text, including all citations and legal analysis, has been reviewed for accuracy and approved by a human" 4. On June 12, 2024, it "decided not to adopt a special rule regarding the use of artificial intelligence in drafting briefs at this time," and stressed that filings "shall be carefully checked for truthfulness and accuracy as the rules already require" 4. Its citation to Federal Rule of Appellate Procedure 6(b)(1)(B) fits poorly; King & Spalding describes that rule as referencing Rule 3(c) and forms for notices of appeal 4. The operative message is general: existing duties already cover AI-assisted work, and "at this time" leaves the door open 4.
Texas state court and local requirements
Mata rests on federal Rule 11 3, so it settles nothing about sanctions in Texas state court, where state procedural rules govern. Confirm which rules apply before you import Mata's analysis.
Where the duties meet: the corrected error
Is one hallucinated citation, promptly corrected, a disciplinable competence failure? The commentary says a lawyer who acts in good faith is not subject to discipline "under those provisions for an isolated inadvertent or unskilled act or omission, tactical error, or error of judgment" 1. That sentence follows a discussion of neglect under paragraph (b), so its reach to paragraph (a) is uncertain.
A lawyer with no process stands on weaker ground, because the failure runs through the practice and is no longer isolated. In Mata, bad faith rested on "acts of conscious avoidance and false and misleading statements to the Court" after the individual respondents "doubled down" 3. In both regimes prompt correction decides the outcome, and you should watch how Texas tribunals weigh it.
Practical guidance
- Verify every authority and quotation against the original before filing. The Fifth Circuit's unadopted standard, "reviewed for accuracy and approved by a human," makes a sound internal rule 4.
- Correct fast and in writing. Mata's bad-faith finding rested on acts of conscious avoidance and false and misleading statements to the court 3. Review your Texas candor obligations too, which apply regardless of any court's sanctions power.
- Make review firm policy. Mata imposed joint firm responsibility under federal Rule 11(c)(1) 3. Test your tool selection, training, and review against the Texas rules on supervising lawyers and nonlawyer assistants.
- Vet confidentiality before you input anything. The committee ties technological competence especially to client confidential information 2, and Opinion 512 asks how lawyers protect it when a tool requires representation information 5. Check data handling and the Texas confidentiality rule first.
- Settle client communication and fees early. The commentary notes "a duty to communicate reasonably with clients; see Rule 1.03" 1, and Opinion 512 asks when to disclose GAI use and what fee is reasonable 5.
- Document training on failure modes. Comment 8 says a lawyer "should engage in continuing study and education" 1.
Bottom line
Neither authority forbids the tool. Mata calls reliable AI assistance not inherently improper 3, and the Fifth Circuit declined a special rule 4. Both demand a lawyer who can catch the tool's errors and correct them openly. A failure to verify generative AI outputs can implicate Rule 1.01 2.