Ethical issues to consider before embracing AI in your law firm

Law firms considering the adoption of any AI tools should ask three questions: what happens to client information, who checks the work, and who is accountable when the system gets it wrong. The ABA and state ethics rules are already applying familiar ethical duties to these new tools. A well-planned AI adoption process can help your firm capture AI’s efficiencies while protecting both clients and the integrity of your work.

AI is already part of everyday legal work. In 8am’s 2026 Legal Industry Report, 69% of surveyed legal professionals said they personally use general-purpose AI tools for work. Yet a surprising 54% of respondents said their firms had provided no training on responsible use of AI and had no plans to do so. In other words, individual AI use is moving faster than firm guidance.

That gap matters. A tool may save time, but lawyers remain responsible for the advice, filings, and client information that comes out of it. Before approving any AI technology, a firm needs to know what happens to the information users enter and who will review the output before it reaches a client or a court.

“Here’s what the courts are telling you about using GenAI: It’s telling you [that you] have to thoroughly review any content you use, that you have ethical obligations in using this,” said Mary Ann Becker, Clinical Professor of Law and Co-Director of Legal Writing at Loyola University Chicago School of Law, summing up the message from the bench. “You have to be keeping up with the technology and its use and its output. It is your duty not to rely on it blindly.”

The professional duties are as old as the legal profession itself: confidentiality, competence, supervision, candor, client communication, reasonable fees, and diligence. Applying those duties in the AI era takes judgment. And responsible policies can depend on the tool, the task, the client, the jurisdiction, and any local court rules that may apply.

What ethics opinions should inform your law firm’s use of AI?

As of today, there is no single AI ethics rule that answers every question for every firm. The ABA’s Formal Opinion 512 offers a useful starting point, but lawyers also have to check their state’s rules and guidance, along with any requirements imposed by the particular court handling a matter.

ABA Formal Opinion 512

Issued in July 2024, ABA Formal Opinion 512 applies existing Model Rules of Professional Conduct to lawyers’ use of generative AI. Its central takeaway is that using AI does not negate a lawyer’s existing duties:

To ensure clients are protected, lawyers using generative artificial intelligence tools must fully consider their applicable ethical obligations, including their duties to provide competent legal representation, to protect client information, to communicate with clients, to supervise their employees and agents, to advance only meritorious claims and contentions, to ensure candor toward the tribunal, and to charge reasonable fees.

The opinion doesn’t require lawyers to tell every client about every use of AI; whether to disclose depends on the circumstances. Confidentiality is stricter. Before entering client information into a self-learning tool, a lawyer needs the client’s informed consent — and boilerplate language isn’t enough.

Of course, ABA opinions are only advisory. The rules adopted in the lawyer’s jurisdiction and any applicable court requirements still control. The ABA opinion is a good starting place; then state guidance applies it in practice.

Texas Opinion 705

Texas Opinion 705, issued in February 2025, warns that lawyers who choose to use AI need a reasonable, current understanding of how their tools work, the limitations, and the risks to client information. That includes reviewing vendor terms, learning about data security, and training lawyers and staff on responsible use.

The opinion draws a clear line between assistance and responsibility:

The best practice here, as with many other efficiency-enhancing tools in the law: AI-generated outputs can be used as a starting point for a lawyer’s work, but must always be carefully analyzed for accuracy and quality.

Texas is also specific about billing. Under an hourly fee agreement, lawyers can charge for actual time spent using and reviewing AI, but not for hours the tool saved, because “the economies associated with the result must inure to the benefit of the client.” This principle isn’t new. The language echoes ABA Formal Opinion 93-379, issued back in 1993.

Florida Opinion 24-1

Florida Ethics Opinion 24-1, issued in January 2024, addresses confidentiality, competent and accurate work, billing, and lawyer advertising. It requires attorneys to check how a tool handles data retention, sharing, and model training, and to establish ways to verify AI-generated work.

The opinion is blunt about verification:

[A] lawyer must verify the accuracy and sufficiency of all research performed by generative AI. The failure to do so can lead to violations of the lawyer’s duties of competence, avoidance of frivolous claims and contentions, candor to the tribunal, and truthfulness to others, in addition to sanctions that may be imposed by a tribunal against the lawyer and the lawyer’s client.

A lawyer also can’t delegate to a tool work that requires the lawyer’s own judgment. And if an AI chatbot communicates with clients or third parties, the bot itself has to follow Florida’s advertising rules and make clear that it is an AI program, not a lawyer.

Florida also says that if a tool won’t disclose confidential information to a third party, client consent isn’t required.

California State Bar guidance and CA SB 574

California’s Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law was updated in May 2026 to address newer developments, including agentic AI. It explains how existing professional duties apply when a tool can plan and carry out tasks, interact with other systems, or act without continuous prompting.

The guidance makes clear that more automation does not mean less lawyer responsibility. Lawyers need to understand what an AI system can do, review and correct its output, and retain control over legal judgment. An agentic system should not make substantive legal decisions, give legal advice, or prepare and file pleadings without meaningful lawyer supervision and review.

On September 30, 2026, California passed state Senate Bill 574 (Chapter 858), the nation's first statutory law regulating the use of generative artificial intelligence by attorneys, arbitrators, and judicial officers, which will go into effect on January 1, 2027. In short, the law bans attorneys from delegating the practice of law to AI and restricts the entry of confidential or personal information into AI tools. It also requires attorneys to verify all citations before filing in court and disclose AI use to the court for all documents submitted. While the law does not require attorneys to disclose AI use in content provided to the public, it does suggest that attorneys “consider whether to disclose” such use.

The California State Bar advises lawyers to assess client communication throughout the representation, considering the tool, its risks, the scope of the work, and the client’s sophistication. Read the updated guidance and the State Bar’s AI ethics resources.

Where to find other state guidance

State guidance continues to develop, so this list is only a starting point. Check your own state bar’s ethics opinions and guidance, then review the rules and standing orders for the courts where you practice. The State Bar of California’s AI resource page links to examples from other jurisdictions across the United States.

Ethical issues to evaluate before implementing AI in your law firm

Ethics opinions offer a framework, but they do not approve a particular tool or use for your firm. Before adoption, ask what information the tool receives, who reviews its output, and whether it can affect client advice or court filings.

Confidentiality: Know where the information goes

A lawyer’s duty of confidentiality covers information relating to a client’s representation, not just privileged communications. Before putting client material into an AI tool, find out whether the provider stores prompts or files, uses them to train models or shares them with others, and how it controls access and deletion. Review the contract and security practices, too. A “no training” setting addresses one concern; it does not answer every question about storage, access, or security.

Phillip Vaden, Partner Manager at InfoTrack and host of The AI Sidebar podcast, says vendors should be able to answer these questions readily.

“Any vendor that’s providing an AI tool...should be able to readily, easily provide you with an explanation of how they’re using your data,” Vaden says. “Are they training on your data? Are they storing your data? What are they doing with it? And they should be able to provide you with a wireframe showing that data flow.”

Use firm-approved tools, share only the information necessary for the task, and determine whether client consent is required under the rules that apply.

Competence: Understand the tool and its limits

Competent use starts with understanding what a tool can do, where it gets its information, and where it can fail. Lawyers do not need to become AI engineers, and they are not required to use AI. You do, however, need enough knowledge to decide whether a particular tool is appropriate for a task and to review its work with your own legal judgment.

“Competence means understanding technology,” says former magistrate judge Ron Hedges.

A fluent answer is not proof of a sound one. AI can summarize a record while missing a key fact or produce a plausible legal analysis that overlooks controlling authority. Before relying on an output, check it against the underlying documents and primary sources.

Supervision: Set rules before the tool is in use

A lawyer’s supervisory duty applies when AI is part of the work, whether the tool is used by a partner, an associate, a paralegal, or another staff member. Hedges offers an example:

[I]f I am a partner in a firm, I am supervising associates, paralegals, and I say in a conference room, ‘Hey, guys. Our firm has just decided to use a generative AI tool. I want you to start experimenting with the tool.’ Is that appropriate supervision? I’m going to suggest to you that the answer to that is no, because you may be putting client confidences and the like into a query or a prompt.

A firm’s AI policy should identify who may use each tool, what information users may enter, which tasks require attorney review, and when and how users should raise a concern. Supervising lawyers also need to train staff and check that the rules are followed.

Automated workflows need the same oversight. Because agentic tools can act without continuous prompting, a lawyer must review the tool’s work before they make a substantive decision, give advice, or file anything.

Candor to the tribunal: Verify before you file

A lawyer who signs or submits a filing is responsible for its contents, whether or not AI helped prepare it.

“All generative AI hallucinates,” says Cheryl Wilson Griffin, VP of Vendor Advisory at Legaltech Hub. “It is not trying to be right. It is looking for plausibility and coherence, not truth. So anytime you’re drafting something with AI, you need to double check all of it.”

Check every authority, quotation, and factual assertion against the original source and the record. Also review the rules and standing orders for the court hearing the matter: some require lawyers to disclose AI use or certify that they verified the filing.

In Lnu v. Blanche, the Ninth Circuit sanctioned two lawyers whose briefs contained nonexistent cases and misrepresentations, and who then failed to be candid about AI’s role. The court imposed fines and six-month suspensions from practice before the Ninth Circuit. It also required future filings by the lawyers and their firm to disclose whether AI was used and certify that the signer reviewed the work and that its citations and quotations referred to existing authority. Those requirements applied to that firm; they are not a nationwide disclosure rule.

When an attorney discovers an AI-generated error, the court’s practical instruction is clear: they “should immediately alert the court and opposing counsel of the error and disclose its source.”

Client consent: When to disclose and how to bill

Decide at the start of an engagement which uses of AI you’ll tell the client about and which require consent, and document both. If a tool could expose client information, get informed consent for that specific use rather than relying on engagement-letter boilerplate. If a client asks whether AI was used, answer directly, and follow any client instructions or contractual limits.

Billing needs the same care. For hourly work, bill for time actually spent, including time spent prompting, reviewing, and correcting AI output. If the firm plans to pass through AI charges, explain that arrangement in advance and ensure the charge is reasonable under the fee agreement and applicable rules. Florida’s opinion recommends telling clients in writing before charging them the actual cost of AI use.

Bias and discriminatory outputs

AI can reflect bias found in its training data, design, or its use of factors that act as proxies for protected traits. A recommendation may sound neutral but still warrant scrutiny, especially when it affects a consequential decision, such as jury selection. Because many tools can’t explain how they reached a particular score, those patterns can be hard to spot.

The risk is concrete in jury selection. ABA Formal Opinion 512 says a lawyer can’t rely on an AI tool’s recommendation to exercise a peremptory challenge if the lawyer knows, or reasonably should know, the challenge would be unlawfully discriminatory. The tool’s score does not shift responsibility away from the lawyer.

Before relying on a tool that ranks or evaluates people, ask what data and factors shape its outputs, what testing the vendor has done, and whether the results could reproduce discriminatory patterns. If the vendor cannot explain how the tool works well enough for the firm to assess those risks, treat that uncertainty as part of the adoption decision. Review the full ABA opinion on discrimination in the jury selection process.

Diligence: Check the tool and keep checking it

Due diligence starts before a tool is approved and continues as the product changes. Review the vendor’s data flow, security controls, retention and deletion practices, and contract terms. Then test the tool on representative work where the firm can compare its output with a known answer. Look for omissions and misleading results, not just obvious errors.

A tool’s built-in review does not replace the firm’s own checks. Reassess the tool when the vendor changes its terms, model, or features, and set a process for pausing its use if the security or quality no longer meets the firm’s standards.

Ensuring your law firm’s AI use is ethical

Before AI enters a client workflow:

• Approve the tool. Review its data practices, security, terms, and limits.
• Define its use. Specify which tasks are allowed, which are off limits, and what information users may enter.
• Assign human review. Name who checks the output and what verification each task requires.
• Train and supervise everyone who uses it. Give attorneys and staff clear rules and a way to raise concerns.
• Check client and court requirements. Address consent, disclosure, billing, and filing rules before the work begins.
• Reassess over time. Track errors, changes to the tool, and incidents; pause use if its risks exceed its value.

Ethical AI use comes down to professional judgment at every stage: choosing the tool, protecting client information, reviewing its work, supervising its users, and meeting duties to clients and courts.

“Lawyers who thrive in the next decade […] are not going to be the ones who resist AI, but they’re also not going to be the ones who outsource their judgment to it, says Neda Mansoorian, Cofounder and CEO, FullProof.ai. “They’re going to be the ones who understand their craft deeply enough to responsibly include an AI second [chair] of some kind in their workflow.”

Sign up for an InfoTrack account today to see how InfoTrack can help your team realize the benefits of AI for your filing workflows.