Ethics and Compliance

Texas Ethics Opinion 705 on Generative AI

What Texas Professional Ethics Committee Opinion 705 says about generative AI: competence, confidentiality, verification, supervision and billing.

AssistCall research teamPublished Reviewed by Faruque Amin, VP of Sales and Marketing7 min read

On this page
  1. The question Texas answered
  2. Competence: no duty to use AI, a duty to understand it
  3. Confidentiality under Rule 1.05
  4. Verification and candor
  5. Supervision: treat AI like a new assistant
  6. Billing
  7. Talking to clients about AI
  8. How Texas compares with ABA Opinion 512
  9. Measure whether a tool is worth it
  10. Example: a Houston firm's first AI policy
  11. A Texas-ready AI policy in five lines
  12. Frequently asked questions
  13. References
Ethics and Compliance: Texas Ethics Opinion 705 on Generative AIArticle coverETHICS AND COMPLIANCETexas Ethics Opinion705 on Generative AIAva for LawyersInsightsEthics and Compliance: Texas Ethics Opinion 705 on Generative AIArticle coverETHICS AND COMPLIANCETexas Ethics Opinion705 on Generative AIAva for Lawyers

Key takeaways

  • Opinion 705 (February 2025) answers what ethical issues the Texas Disciplinary Rules raise when lawyers use generative AI[1].
  • Lawyers need not use AI, but those who do need a reasonable and current understanding of it and must verify its output[1].
  • Do not input confidential information unless reasonably satisfied the tool will not reveal it to others; otherwise consult the client and get consent[1].
  • Bill actual time, not time saved; per-use tool costs may be passed on only by agreement and only as actually incurred[1].

The question Texas answered

In February 2025, the Professional Ethics Committee for the State Bar of Texas issued Opinion 705, answering a single question: what ethical issues are raised under the Texas Disciplinary Rules of Professional Conduct by a lawyer's use of generative artificial intelligence in the practice of law[1]. Texas had 13,833 offices of lawyers in 2023, the third most of any state[2], and the opinion now frames how those firms are expected to evaluate AI tools.

The committee describes its opinion as a snapshot that does not address every issue, and acknowledges that some of the issues it raises may soon be resolved or mooted by changes in the technology[1]. Read it as a set of principles rather than a list of approved or banned products.

Competence: no duty to use AI, a duty to understand it

Opinion 705 does not require lawyers to use generative AI. Lawyers who do use it must understand the risks, including inaccurate output and exposure of data, and need a reasonable and current understanding of the technology, consistent with the comment to Rule 1.01 urging lawyers to become and remain proficient and competent[1].

Most professionals are not there yet. In Thomson Reuters research, 89% of professionals said they see use cases for generative AI, but 64% had received no training on using it for their work[3]. In Texas terms, that gap is a competence problem waiting to happen.

Professionals on generative AI: opportunity versus training
Professionals on generative AI: opportunity versus trainingComparison. See use cases for generative AI: 89%; Have received no training on it for their work: 64%.89%See use cases for generative AISee use cases for generative AI: 89%64%Have received no training on it for theirworkHave received no training on it for their work: 64%Professionals on generative AI: opportunity versus trainingComparison. See use cases for generative AI: 89%; Have received no training on it for their work: 64%.89%See use cases for generative AISee use cases for generative AI: 89%64%Have received no training on it fortheir workHave received no training on it for their work: 64%

Source: Thomson Reuters, 2025[3]

Show the data
ItemValue
See use cases for generative AI89%
Have received no training on it for their work64%
  • Train before access: no one uses a tool on client work until they have completed the firm's short training.
  • Teach failure modes, not just features: invented citations, confident errors and outdated law.
  • Refresh training when the tool or its settings change.

Confidentiality under Rule 1.05

Entering client information into an AI tool can implicate Rule 1.05, particularly prompts that disclose facts or legal theories. The opinion explains that some tools store and incorporate user inputs into their datasets, creating a risk that confidential information surfaces for other users. A lawyer should be reasonably satisfied that the program will not reveal confidential information to others; without that assurance, the lawyer should not input confidential information without consulting the client and obtaining consent[1].

  1. Read the data terms

    Find out whether inputs are stored, used for training or shared, and for how long.

  2. Record your conclusion

    Write down why the firm is reasonably satisfied, or not, and keep it with the tool approval.

  3. Strip identifiers

    Where a task does not need names or unique facts, remove them before prompting.

  4. Consult when unsure

    If the firm cannot be reasonably satisfied, explain the risk to the client and obtain consent before using their information.

Verification and candor

Opinion 705 says lawyers must verify AI output for accuracy and quality, and that failing to do so can violate several rules, including the duty of candor toward the tribunal. Lawyers are responsible for the work product they submit regardless of who, or what, did the original research[1]. The opinion points to Mata v. Avianca, the 2023 federal case in New York where lawyers were sanctioned for a brief citing non-existent opinions invented by a general-purpose AI chatbot, as the most famous example of improper use[4].

The risk is not rare. A Stanford study found general-purpose chatbots hallucinated on between 58% and 82% of legal queries[5]. Every citation in an AI-assisted filing should be pulled and read by a person before it is filed.

Supervision: treat AI like a new assistant

The opinion compares AI to nonlawyer assistance under Rule 5.03. Lawyers who rely on AI risk many of the same perils as those who rely on inexperienced or overconfident nonlawyer assistants[1]. The comparison is useful in practice: you would not let a new assistant send advice to a client or file a document without review, and the same limits fit an AI tool.

For client-facing tools, the analogy is even closer. An AI that answers calls or chats with prospective clients is doing the work of a receptionist or intake specialist, so it should follow the same script limits: identify itself honestly, collect facts, give no legal advice and pass legal questions to a lawyer.

Billing

Lawyers may bill for the actual time spent using an AI program, including reviewing its output, but may not bill hourly for time the AI saved; efficiency gains belong to the client. Per-use AI costs may be passed on to clients if the arrangement is agreed, and recovery is limited to expenses actually incurred and paid[1].

  • Record real time on AI-assisted tasks, including prompting and review.
  • If a tool has a per-use cost you want to pass on, put it in the fee agreement first.
  • Never mark up a tool cost as an expense; charge what was actually paid.

Talking to clients about AI

Opinion 705 says lawyers should consider informing clients about the risks of AI use and may need consent when confidential information is involved, noting that other jurisdictions recommend disclosure or informed consent and that Texas lawyers should apply a similar analysis[1]. In practice, three situations call for a conversation rather than a line in the engagement letter.

  • The firm wants to use a tool with the client's confidential information and is not fully satisfied with its data handling.
  • AI will do a large share of a task the client is paying for, such as reviewing a document production.
  • The client has asked whether the firm uses AI, or has its own policy restricting outside counsel's use of it.

Keep the conversation short and specific: which tool, what it will be used for, how the firm checks the output and how it affects the bill. Note the discussion and any consent in the matter file.

How Texas compares with ABA Opinion 512

Texas Opinion 705 and ABA Formal Opinion 512 side by side.
TopicTexas Opinion 705ABA Formal Opinion 512
StatusTexas committee opinion, February 2025ABA committee opinion, July 2024
Confidentiality standardBe reasonably satisfied the tool will not reveal information; otherwise consult and get consentInformed consent before using self-learning tools; boilerplate is not enough
VerificationRequired; responsible regardless of who or what did the researchRequired; independent verification of output
SupervisionAI compared to nonlawyer assistants under Rule 5.03Governance, policies and training under Rules 5.1 and 5.3
FeesActual time only; per-use costs by agreement at actual costActual time only; no charge for learning a tool
Client communicationConsider informing clients; consent may be neededDisclosure may be required depending on use and impact

The two are consistent in direction. Texas frames confidentiality around whether the lawyer is reasonably satisfied with the tool, while the ABA focuses on informed consent for self-learning tools. A firm that meets the ABA's consent standard and documents why it is satisfied with each tool's data handling will meet both.

Measure whether a tool is worth it

Competence includes knowing whether a tool helps. Only 20% of legal professionals surveyed by Thomson Reuters said they measure the return on investment of generative AI[6]. Before adopting a tool firm-wide, run a short trial on closed matters and record time spent, error rates found in review and how often the output was usable. That evidence supports both the decision and the billing approach.

Example: a Houston firm's first AI policy

Example: a fictional Houston litigation boutique wants to use a research assistant and a drafting tool. Following Opinion 705, the firm reads both vendors' data terms, concludes it is reasonably satisfied with the research tool but not with the free drafting tool, and bans the latter for client work. It trains every lawyer and paralegal for an hour, adds a citation-check step to its filing checklist, and updates its fee agreement to say that AI-assisted time is billed as actual time and that no tool costs are passed on. Six months later it reviews the trial data and expands the research tool to the whole firm.

A Texas-ready AI policy in five lines

  1. Approved tools only, each with a documented data review.
  2. No confidential information unless the firm is reasonably satisfied or the client consents.
  3. Every output verified by a person before use; every citation read before filing.
  4. Tools supervised like nonlawyer assistants, with client-facing scripts approved by a lawyer.
  5. Bill actual time; pass on tool costs only by agreement and at actual cost.

For other states' approaches, see California's guidance, Florida Opinion 24-1 and New Jersey and Pennsylvania.

Frequently asked questions

What is Texas Opinion 705?

A February 2025 opinion of the Professional Ethics Committee for the State Bar of Texas on the ethical issues raised by lawyers' use of generative AI under the Texas Disciplinary Rules of Professional Conduct.

Does Texas require lawyers to use AI?

No. Opinion 705 says lawyers need not use generative AI, but those who do must understand its risks and keep a reasonable and current understanding of the technology.

Can Texas lawyers put client information into AI tools?

Only if reasonably satisfied the tool will not reveal it to others. Otherwise, the lawyer should consult the client and obtain consent first.

How should Texas lawyers bill for AI-assisted work?

Bill the actual time spent, including reviewing output, not time saved. Per-use tool costs can be passed on only if agreed and only as actually incurred and paid.

References

  1. Texas Center for Legal Ethics (Professional Ethics Committee for the State Bar of Texas) (2025). Opinion 705 (opens in a new tab) โ†‘

  2. U.S. Census Bureau (2023). County Business Patterns 2023: state file (cbp23st) (opens in a new tab) โ†‘

  3. Thomson Reuters (2025). Generative AI Adoption Nearly Doubles as Professional Services Reach Crossroads (opens in a new tab) โ†‘

  4. Texas Center for Legal Ethics (Professional Ethics Committee for the State Bar of Texas) (2025). Opinion 705 (opens in a new tab) โ†‘

  5. Stanford Institute for Human-Centered Artificial Intelligence (HAI) (2024). AI on Trial: Legal Models Hallucinate in 1 out of 6 (or More) Benchmarking Queries (opens in a new tab) โ†‘

  6. Thomson Reuters Institute (2025). 2025 GenAI report: Executive summary for legal professionals (opens in a new tab) โ†‘

Book a demo

See Ava handle your calls.

Book a short demo. Weโ€™ll walk through your calls and workflow, and agree what Ava should take care of.

Book a demo