AI Ethics in Law: Professional Duties, Risks, and Best Practices for Lawyers

More lawyers are using AI for legal research, document review, and drafting. According to the latest Legal Trends Report, 79% of legal professionals now use AI in their firms. But as AI becomes embedded in everyday legal work, one question is becoming impossible to ignore: What do professional ethics rules actually require?

AI ethics in law isn’t a single rule. For lawyers, it covers specific obligations that govern competence, confidentiality, supervision, and candor. Using AI doesn’t suspend those obligations; knowing where they apply is what keeps your practice protected.

In this guide, we’ll walk through the ethics rules covering AI use in legal practice, how the American Bar Association (ABA) Formal Ethics Opinion 512 and the ABA’s Model Rules of Professional Conduct map to practice decisions, what court disclosure requirements demand, an evolving state-by-state overview of bar guidance, and practical steps for integrating AI into your practice responsibly.

What are the ethics rules for using AI in law? (2026)

Lawyers using AI must comply with several ABA Model Rules of Professional Conduct (Model Rules), court procedural rules, and any court-specific disclosure orders in their jurisdiction. The most relevant Model Rules are 1.1 (competence), 1.4 (client communication), 1.5 (reasonable fees), 1.6 (confidentiality), 3.3 (candor toward the tribunal), 5.1 and 5.3 (supervision), and 5.5 (unauthorized practice of law).

Certification duties under court procedural rules also apply to AI-assisted filings. Lawyers must certify accuracy and legal support whether or not AI helped draft the document. ABA Formal Opinion 512, issued in 2024, pulls these obligations together into a single framework and adds a vendor due diligence requirement. The sections below walk through each rule in detail.

Legal ethics and AI: How the ABA Model Rules apply to AI use

The ABA Model Rules don’t reference AI explicitly, but several rules speak directly to how lawyers must approach it. Here’s what each one requires.

Rule 1.1 – Competence

Competence requires lawyers to keep abreast of changes in the law and its practice, including the benefits and risks associated with relevant technology. In practice, lawyers who use AI must understand how those tools work—including their limitations—well enough to evaluate their outputs critically. The rule also requires lawyers to consider benefits, not just risks. Benefits of AI for lawyers may include greater efficiency, speed, and more comprehensive topic coverage, all of which matter to clients. Lawyers can’t disregard those benefits simply to avoid risk. The standard is a cost-benefit analysis.

Risk to watch: AI models can generate confident, well-structured responses that are factually wrong (a phenomenon known as AI hallucinations) and bias in training data can produce outputs that underrepresent smaller jurisdictions or recent changes in the law. Treat AI outputs as drafts, verify citations against primary sources, and prefer tools built specifically for legal use with jurisdiction-aware sourcing.

Rule 1.4 – Client Communication

When AI materially affects the services you provide, the fees you charge, the confidentiality of their information, or any important decision in their matter, clients need to know. Rule 1.4 requires lawyers to keep clients reasonably informed and to explain matters as needed for them to make informed decisions. Depending on the situation, firms should consider including an AI clause in engagement letters explaining which tools may be used and how client data is protected. For a practical template, see our AI disclosure guide.

Rule 1.5 – Reasonable fees

Fees must be reasonable. When AI reduces the time a task takes, billing at your hourly rate as though it didn’t is not permitted. Don’t bill for time not spent. Instead, consider fee structures that reflect the value of the work rather than hours logged—flat fees, tiered pricing, or subscription retainers can all work well in an AI-assisted practice.

Both California and Virginia recently issued updated ethics guidance stating that AI usage can be considered when determining fees, and not just in the calculation of hourly rate amounts. California suggests lawyers “build the time savings created by AI use into their competitive rates.” Virginia’s ethics opinion states that, “Rule 1.5 does not equate reduced time with proportionally reduced fees.” The opinion’s discussion stresses that the time and labor required is just one factor to consider when setting a reasonable, non-hourly fee structure.

Whichever structure you choose, explain any AI-related costs to clients upfront. Some firms add a line item for AI tool costs where available and those are passed through; others absorb them. Either approach is defensible if the total fee is reasonable and the client is informed.

Rule 1.6 – Confidentiality

Lawyers must take reasonable precautions to prevent inadvertent or unauthorized disclosure of client information. Before inputting any client data into an AI tool, lawyers must understand how that tool stores, processes, and shares information. Many general-purpose AI tools, including free or consumer-grade products, are not designed to meet legal confidentiality standards.

Risk to watch: General-purpose tools like ChatGPT may allow conversation data to be used for model training and be considered by clients an unauthorized disclosure. Courts have ruled that communications made through tools with no reasonable expectation of privacy may not meet the duty of confidentiality, regardless of who initiated them. A practical rule of thumb: if you wouldn’t email the information to an unvetted third party without a binding confidentiality agreement, don’t enter it into an AI tool without first understanding its data handling practices.

Rule 3.3 – Candor toward the tribunal

Lawyers have a duty not to make false statements of law or fact to a court. Submitting AI-generated legal research without verification, including citations that may be fabricated, can directly violate this rule.

Risk to watch: Courts have already sanctioned lawyers for filing briefs containing AI-generated but nonexistent case citations. Verify every citation against primary sources before filing.

Rules 5.1 / 5.3 – Supervision

Partners and supervising lawyers are responsible for ensuring that subordinates, and the non-lawyer tools and systems they use, comply with professional obligations. This means AI outputs must be reviewed, not merely accepted. A lawyer can’t delegate professional judgment to an algorithm, and relying on AI-generated work product without meaningful review likely falls short of this standard.

Risk to watch: Accountability for AI errors sits with the lawyer, not the tool. Legislatures are moving to make that explicit. See the sanctions section below for the current state of statutory accountability.

Rule 5.5 – Unauthorized practice of law

Rule 5.5 prohibits the unauthorized practice of law. When AI chatbots deployed by law firms or legal tech platforms provide responses that could be misconstrued as legal advice without adequate attorney oversight, they can raise UPL concerns or attorney-client relationship misunderstandings. Lawyers who deploy client-facing AI tools should ensure those tools are supervised, clearly scoped, and do not substitute for attorney judgment on matters requiring professional analysis.

For a deeper look at how to stay compliant with these rules as AI evolves, see our companion guide on AI compliance for lawyers.

ABA Formal Opinion 512: pulling it together

In 2024, the ABA Standing Committee on Ethics and Professional Responsibility issued Formal Ethics Opinion 512, its first comprehensive guidance on generative AI. Opinion 512 doesn’t create new rules; it applies the Model Rules above to AI use and draws out one obligation that the rules alone don’t make explicit: vendor due diligence.

Before deploying any AI tool in client matters, lawyers should review the vendor’s terms of service, data retention policies, and security certifications. When vetting a tool, look for evidence on how it complies with your ethical obligations. Seek documentation of technological and organizational controls, like zero data retention agreements confirmed in writing, SOC 2 Type II certification, encryption in transit and at rest, role-based access controls, audit logs for all AI interactions, and terms of service that don’t permit training on client data.

The full opinion is available on the ABA’s website. Every practicing lawyer using AI should be familiar with it.

What happens when lawyers get AI ethics wrong: sanctions and case law

The ethics obligations above aren’t theoretical. Courts have increasingly imposed sanctions on lawyers who submitted AI-generated work product without adequate review, and legislatures are moving to codify verification duties into law. Here’s what the enforcement record looks like so far.

• Mata v. Avianca (S.D.N.Y. 2023) is the most cited example. Attorneys submitted a brief containing multiple citations to cases that didn’t exist—fabricated by an AI tool and submitted without verification. The court imposed sanctions, required the attorneys to notify the judges whose names appeared in the nonexistent opinions, and ordered them to pay costs. The case put the legal profession on notice that AI hallucinations in court filings carry real consequences.
• United States v. Heppner illustrates the confidentiality risk. A defendant used a free public AI chatbot to generate defense strategies. The court ruled those communications unprotected because the tool offered no meaningful confidentiality controls. While the case involved a client rather than a lawyer, it shows how much turns on the data handling practices of the tool being used.
• Lnu, et al. v. Blanche highlights the candor risk. A federal appeals court panel ordered sanctions, not for using generative AI, but for filing briefs containing hallucinations, and due to their lack of candor in revealing that these inaccuracies were the result of AI hallucinations. The sanctionable misconduct, according to the court, was not that an AI-generated error occurred in a brief, but that the lawyers knowingly or recklessly made false statements to the court when asked to show cause.
• Withers v. City of Aberdeen is a recent sanctions order that included the judge questioning attorneys about their CLE education on AI risks and whether the law firm had an AI-usage policy. The court found that an admitted failure to independently verify AI outputs in violation of the firm’s written policy requiring that verification was reason to impose a greater fine amount.

Taken together, these developments point in one direction: The standard for AI use in legal practice is being defined in real time, and “I relied on the tool” is not a defense.

State bar ethics guidance on AI

Bar associations are moving at different speeds, but most have issued or are developing formal guidance on AI use. Below is a snapshot of key jurisdictions. For a detailed 50-state survey, see Justia’s attorney ethics and AI roundup.

This list isn’t exhaustive. Check your state bar’s ethics resources directly for the most current guidance. 

Court AI disclosure and certification requirements

Courts are increasingly requiring lawyers to certify their AI use in filings. What’s required varies by jurisdiction and is evolving quickly, so always check local rules before filing.

The general pattern across courts that have issued orders requires attorneys to either certify that no portion of a filing was drafted by AI, or disclose that AI was used and confirm that a human reviewed and verified all citations and content for accuracy. Some orders are narrow, covering only generative AI used in drafting. Others are broader, applying to any AI use in preparing a filing.

One active example is the standing order from Senior District Judge Michael Baylson of the Eastern District of Pennsylvania, which requires any attorney who used AI in preparing a filing to include a clear factual statement disclosing that use and to certify that every citation to law or the record has been verified for accuracy.

As of June 2026, the Ropes & Gray AI court order tracker records 681 cases and court rules on AI use across federal and state courts. Check it alongside your district’s local rules page for the most current requirements in your jurisdiction.

Even where no formal order exists, documenting your AI use—which tools you used, for what tasks, and how you reviewed the output—gives you a clear record to produce if a court ever asks.

Courts may require certification of citations, even when AI has not been used on the filing. Starting June 15, 2026, lawyers who sign filings submitted in state courts are representing that the legal authorities identified within exist and are accurately cited. This update to Florida’s Rules of General Practice and Judicial Administration §2.515(d)(2) does not reference AI, but places responsibility for any hallucinated citations on the lawyer signer. This brings Florida’s state court rules closer in alignment to U.S. federal court rules.

FRCP 11 and AI-generated filings

U.S. Federal Rule of Civil Procedure 11 requires counsel to certify that every filing is, to the best of their knowledge, supported by existing law or a nonfrivolous argument for modifying that law, and that factual contentions have or will likely show evidentiary support. That certification applies regardless of whether a human or an AI drafted the document.

Lawyers who submit AI-generated filings without verifying citations and legal representations risk Rule 11 sanctions. Before any filing that involves AI assistance, verify that each citation exists, pull the actual holding, and make sure the representation in your brief matches the source.

How to ethically integrate AI into legal workflows

Knowing the rules is only part of the work. Here’s what responsible AI adoption looks like in practice, mapped to the obligations above.

1. Establish a firm AI policy (Rules 5.1, 5.3, 1.1). Document which tools are approved, what data can be inputted, and how outputs must be reviewed. Having it in writing protects your clients and gives your team a consistent foundation. See our law firm AI policy template for a starting point.
2. Apply human oversight where it matters (Rules 5.1, 5.3, 3.3). AI outputs shouldn’t reach a client or a court without a lawyer’s review. Use the AI-generated content checklist for lawyers to make sure nothing slips through.
3. Vet tools for confidentiality compliance (Rule 1.6). Check data retention and privacy policies before using any tool with client information. Prefer tools designed for legal use with enterprise-grade protections built in.
4. Document your AI use (Rules 5.1, 5.3). Keep a record of which tools you used, for what tasks, and how you reviewed the output. Some jurisdictions already require disclosure of AI use in filings, and good documentation is your protection if questions arise later.
5. Train your team (Rules 5.1, 5.3). Your supervision obligations extend to how staff and associates use AI. A short onboarding session on your firm’s AI policies and a tool’s limitations goes a long way toward reducing risk.
6. Stay current on bar guidance (Rule 1.1). This area is moving fast. Subscribe to updates from your state bar and the ABA’s Center for Professional Responsibility, and revisit your AI policy at least once a year.

Staying on the right side of AI ethics

The professional duties that govern AI use—competence, confidentiality, supervision, and candor—have governed lawyer conduct under the ABA Model Rules since 1983. AI is a new tool; the standard it’s held to isn’t new.

What’s changed is the surface area for error. AI can produce convincing work product that is wrong, leak confidential data through an insecure tool, or create a paper trail of unreviewed outputs that exposes a lawyer to sanctions. The obligations that protect against those risks have always existed. Meeting them now requires understanding how AI fits into each one.

Stay informed, apply meaningful oversight, protect client confidentiality, and keep your judgment at the center of every decision. That’s what responsible AI adoption looks like in practice, and what the rules have always required.

Put AI to work on your cases with confidence. Clio Work delivers cited research, document analysis, and case strategy grounded in verified law and the full context of your matters. Book a demo today.