AI Regulation Tracker / Legal profession and ethics
Texas Ethics Opinion 705 Sets the Bar for Lawyers Using Generative AI
Issued in February 2025, it is guidance rather than a rule, but it is exacting. The Professional Ethics Committee for the State Bar of Texas told lawyers that if they use generative AI, they must actually understand it, keep client confidences out of leaky tools, check every output, and never bill for hours they did not spend.
A year after Florida went first, Texas took its turn, and the result is one of the more quotable state-bar statements on lawyer AI use. In February 2025 the Professional Ethics Committee for the State Bar of Texas issued Opinion 705, addressing the ethical issues raised when lawyers use generative AI in practice. Like Florida's opinion, it is advisory guidance rather than a rule, and like Florida's, it does not ban the technology. What it does is spell out, in blunt language, the duties that come attached.
You have to understand the tool
Texas leads with competence, and it sets a real threshold. The opinion states that "if a lawyer opts to use a generative AI tool in the practice of law, the lawyer must have a reasonable and current understanding of the technology." The word current is doing work there. This technology changes fast, and a one-time briefing from two years ago does not satisfy the duty. If you are going to put AI into client work, you are expected to keep up with how it behaves, what it can and cannot do, and where it fails. Ignorance of the tool is not a defense.
Keep client confidences out of leaky systems
On confidentiality, the committee is careful. Before a lawyer inputs sensitive information into a generative AI system, the opinion says "the lawyer should be reasonably satisfied that the program will not reveal confidential information to others." That is the crux of the confidentiality problem with third-party AI. Some tools retain, share, or train on your inputs, and a lawyer who pastes client confidences into a system like that may be handing them to the vendor and beyond. The duty is to check first and be satisfied the tool will hold the line, not to assume it does.
Never file what you did not verify
This is the duty that has already produced sanctions across the country, and Texas states it plainly: "lawyers cannot blindly rely upon or use answers given by generative AI tools." The opinion is explicit about why, warning against the notorious hallucination problem where these tools generate convincing but entirely fabricated legal citations and information. The lesson from the sanctions cases is exactly this. AI-generated authority looks real. It has proper formatting, plausible names, believable reporter cites, and some of it is invented. The lawyer who files it without checking owns the fabrication. Verification is not optional polish; it is the duty.
Do not bill AI speed as billed time
On fees, the committee draws a clean line. Time savings from AI efficiency cannot be converted into billed hours. As the opinion puts it, "a lawyer who has undertaken to bill on an hourly basis is never justified in charging a client for hours not actually expended." If AI compresses eight hours of work into two, you bill two. The client gets the benefit of the efficiency, not a bill for phantom time. It is the oldest billing rule in the book, applied to a new efficiency, and the committee did not carve out an AI exception to it.
Why this matters beyond Texas
Opinion 705 is advisory. It interprets the Texas Disciplinary Rules of Professional Conduct rather than adding a new binding rule, so its formal status is guidance. But the rules it applies, competence, confidentiality, candor, and reasonable fees, are binding, and a Texas lawyer who disregards the opinion is still exposed under them. The bigger signal is convergence. Read Texas next to Florida and the pattern is unmistakable: understand the tool, protect confidences, verify everything, bill honestly. Different states, different rule numbers, same four duties. For any US lawyer, that consistency is the actual takeaway. You do not need to memorize every state's opinion. You need to run every one of these four checks, every time, on every AI-assisted matter.
Questions professionals are asking
Does Texas Opinion 705 prohibit lawyers from using AI?
No. It permits generative AI in legal practice and instead sets conditions: the lawyer must have a reasonable and current understanding of the technology, protect client confidentiality, independently verify output, and bill honestly. It is guidance on how to use AI ethically, not a ban.
What level of AI knowledge does the opinion require?
A reasonable and current understanding. The opinion states that a lawyer who opts to use a generative AI tool must have a reasonable and current understanding of the technology. Because the technology evolves quickly, a stale or one-time understanding does not satisfy the competence duty.
Can I put client information into a generative AI tool?
Only after you are reasonably satisfied the program will not reveal confidential information to others. That means checking how the tool retains, shares, or trains on inputs before entering anything sensitive, rather than assuming it keeps confidences.
Can I bill clients for the time AI saved me?
No. The opinion states that a lawyer billing on an hourly basis is never justified in charging for hours not actually expended. If AI makes the work faster, the client is billed for the time actually spent, not the time the work would have taken without AI.
RELATED BRIEFINGS
Browse the full AI Regulation News tracker
Informational analysis for working professionals, not legal advice. Confirm how any ethics opinion or rule applies to your practice with your bar's ethics counsel in the relevant jurisdiction.