AI and Legal Ethics: Important Considerations for Attorneys Using AI

AI, or artificial intelligence, refers to systems that perform tasks traditionally requiring human intelligence, including reasoning, pattern recognition, and learning from experience. AI has become part of modern legal practice, bringing both meaningful opportunities and important ethical responsibilities.

Many legal professionals use generative AI(1) for a broad range of tasks—from summarizing documents and generating charts of parties to drafting pleadings, performing legal analysis, and organizing evidence.

We are all familiar with calculators and computers—machines that can execute predefined instructions and rely entirely on the user’s own knowledge and reasoning. If a student has never learned multiplication, a calculator cannot teach them how to solve an algebra problem.

AI is different. Unlike traditional technology that follows fixed instructions, AI can analyze patterns, learn from data, and generate human-like responses. A generative AI tool may produce competent writing or analysis even when the user could not have completed that work on their own. And that is where much concern about AI begins: the fear that AI will replace the attorney.

That concern, however, misses a critical point. AI cannot replace the human judgment, nuance, ethical responsibility, strategic thinking, and lived experience that define effective lawyering. A machine simply cannot fully appreciate how a particular argument may affect settlement negotiations, influence opposing counsel, or resonate with a jury.

Still, while legal professionals should set aside their worry about being replaced by AI, there are legitimate concerns, including legal ethical duties, surrounding AI adoption that deserve careful attention.

Professional responsibility in the age of AI

Clients—and society in general—expect professionals to use modern technology, and this increasingly includes AI. Just as clients expect attorneys to use email, text messaging, and websites, they will also expect lawyers to understand and appropriately use the advanced technologies available today.

These market expectations, along with the Model Rule of Professional Conduct 1.1 Competence, obligate legal professionals to understand the technologies they use to deliver legal services.

Additionally, with Model Rule of Professional Conduct 1.4 Communication, attorneys are required to reasonably consult with the client about how his or her objectives will be accomplished. The rule mandates informing the client and obtaining his or her consent when using AI to assist in legal services.

When AI gets it wrong

AI can make mistakes. The information it produces may be outdated, biased, misleading, or fabricated—a phenomenon commonly referred to as an “AI hallucination.” These risks are often heightened when using open or public AI models.

One of the earliest and most widely discussed examples is the 2023 case Mata v. Avianca, in which an attorney used ChatGPT to submit fake citations and nonexistent cases, then failed to correct the issue when questioned by the court. The result was sanctions(2) and widespread professional scrutiny. Similar incidents have continued to surface in the years since.

Courts and professional organizations have treated these situations as breaches of the duties of supervision (Model Rule of Professional Conduct 5.1 and 5.3) and candor toward the tribunal (Model Rule of Professional Conduct 3.3).

Some courts are also implementing rules addressing AI opacity, where reliance on questionable AI-generated evidence or filings may result in adverse inferences against the presenting party. In practical terms, if an attorney cannot explain missing, altered, or unreliable information, courts may presume the evidence would have been unfavorable to the client.

The broader concern is not difficult to see. Misleading, biased, outdated, or fabricated information has the potential to confuse or improperly influence litigants, juries, the public, or precedent. The public should be able to trust that attorneys are working from facts, accuracy, and sound reasoning rather than information that may be inaccurate or manipulated.(3) This responsibility remains unchanged in the age of AI.

Client confidentiality and privilege

Client information, matter data, or work product that is uploaded into an open or public AI platform constitutes a per se voluntary disclosure to a third party(4) and can potentially destroy any reasonable expectation of confidentiality.

When a legal professional—or even a client—uploads sensitive information into an open AI platform, it is, in many ways, akin to being scattered publicly with no meaningful ability to retrieve or contain it.

Unsecured AI systems can create significant risks involving personal, financial, medical, and other identifying information. These disclosures violate not only Model Rule of Professional Conduct 1.6 Confidentiality of Information, but also may violate state, federal, and international privacy laws, including HIPAA and other data protection laws.

Importantly, many of these concerns are tied to open or public AI tools—not platforms developed specifically for the high standards of legal work.

The AI efficiency paradox

Many legal professionals today are grappling with how AI-driven efficiency can affect billing and profitability. Tasks that once required hours may now take only minutes with AI assistance. This creates an understandable tension: the more efficiently attorneys work, the harder it may feel to justify traditional billing structures. Questions will naturally arise regarding what constitutes a “reasonable fee” when using AI can significantly reduce the time required to complete work.

In response, some firms are considering flat-fee or value-based billing models. Yet even those approaches raise additional ethical and practical questions. For example, the ABA Statutory Model (ABA Model Rule 1.5) has deemed flat fees potentially unreasonable if AI has drastically reduced the amount of effort required. The profession is actively working through what fair, transparent, and client-centered billing should look like in an AI-assisted environment.

Five steps for using AI ethically and responsibly

The good news is that these risks are manageable. With thoughtful safeguards, clear policies, and sound professional judgment, legal professionals can use AI responsibly while protecting competence, confidentiality, candor, and client trust.

The following five steps offer a practical framework for integrating AI into legal practice in a way that is ethical, defensible, and useful:

1. Build AI competence

You do not need to be an AI expert, but you should develop a reasonable understanding of the capabilities and limitations of the AI tools your firm uses. That education should extend to your entire team. Ongoing training helps to ensure everyone understands not only how to use AI effectively, but also where caution and human oversight remain essential.

It can also be valuable to designate at least one especially tech-savvy or AI-certified team member to serve as an internal resource during the firm’s adoption process and help guide ongoing education as the technology evolves.

2. Establish clear AI usage guidelines

Today, every firm should create and maintain a clear, accessible Acceptable Use Policy (AUP) governing AI use across the organization. By creating a clear, accessible AUP, law firms can also help reduce the risks associated with shadow AI—staff using AI tools without formal approval and oversight.

That policy should address applicable international, federal, and state laws, confidentiality obligations, approved tools, acceptable use, and supervisory expectations. Firms should also train team members on the policy, require attestation, and regularly revisit the policy as AI technology and regulations evolve.

3. Use trusted and secure legal AI

Carefully vet the proprietary AI tools your firm uses and restrict or prohibit the use of any open AI models. Whenever possible, firms should choose AI systems that prioritize security and are built specifically for the legal industry, with features designed to minimize the exposure of confidential information and boost reliability through trusted legal databases.

4. Always maintain human oversight

Treat AI as an inexperienced assistant, never a final decision-maker. Avoid falling victim to automation bias, or the tendency to place undue trust in machine-generated outputs. Human-in-the-loop (HITL) features are also essential to consider. With HITL involvement in legal tasks, a qualified attorney can verify AI output, helping to further ensure that facts, legal propositions, citations, conclusions, and recommendations are accurate. Nonetheless, your own professional judgment and expertise must always serve as the final authority.

5. Be transparent about AI use

Consistent with the Model Rule of Professional Conduct 1.4, firms should communicate transparently with their clients on how AI may be used within the representation. This may include clearly explaining the advantages of using AI for efficiency, organization, and research—all of which can help legal professionals devote more time to high-value tasks like case strategy and client support.

Firms should also explain the safeguards in place to help protect confidentiality and maintain professional standards. Additionally, firms may wish to obtain explicit written consent and consider providing clients with a list of tasks where AI may be utilized, so clients may accept or decline. Transparent billing practices may also help clients better understand both the impact of AI and the continued value of attorney expertise.

Build an ethical foundation for legal AI use

AI is neither a substitute for attorneys nor a threat to the profession when used thoughtfully and responsibly. The real challenge for legal professionals is not whether to engage with AI, but how to do so in a way that protects clients, preserves trust, and upholds the standards of the profession.

By approaching AI with competence, caution, and transparency, attorneys can embrace innovation while remaining firmly anchored to the ethical duties that define the practice of law.

For additional insight into why now is the time to thoughtfully implement AI—and the legal ethics considerations that come with it—read our thought paper, “The Legal AI “Phony War” and the Real Disruption That Lies Ahead.” It explores why this moment matters and how law firms can approach AI strategically before the landscape shifts more dramatically.

ENDNOTES:

1. Generative AI: a type of artificial intelligence that creates content (text, images, audio, code, models) by learning patterns from large datasets. Generative AI uses models to deliver human-like output in response to user prompts.

2. Most severe sanctions: District 109,700 (2026); 6th Circuit – Whiting v. City of Athens ($116,315), SD of Florida – Byoplanet International, LLC v. Johansson ($85,567.75). of Oregon $

3. Consider Time Magazine’s OJ Simpson cover which altered his skin color: https://www.ebsco.com/research-starters/communication-and-mass-media/time-magazine-cover-uses-altered-o-j-simpson-photo.

4. US v. Heppner, 25 Cr. 503 (JSR), 2026 BL 52143 (SDNY Feb 17, 2026).