Artificial Intelligence, Attorney Ethics, and Copyright: Navigating the New Era

Artificial intelligence is no longer an abstraction on the horizon of legal practice. It is already here — in the research tools attorneys use, in the drafting assistants built into modern word processors, and in the client-facing platforms that answer preliminary legal questions at scale. The question facing every practicing attorney is no longer whether AI will affect the profession, but how to engage it competently, ethically, and lawfully.

This article, drawn from a presentation delivered to the American Inns of Court, walks through the core ethical obligations triggered by AI adoption, explores the copyright questions raised by AI-generated content, and draws lessons from real-world case studies — including cases where AI use went very wrong.

---

The State of AI Adoption in the Legal Profession

AI adoption among attorneys has accelerated dramatically. Surveys consistently show that a majority of legal professionals have used generative AI tools in their practice — for research, summarization, drafting, and client communication. Large language models such as ChatGPT, Claude, Gemini, and legal-specific platforms like Harvey, CoCounsel, and Lexis+ AI have moved from novelty to mainstream.

The legal profession's relationship with technology is not new. Attorneys have adapted to typewriters, fax machines, computerized research, e-discovery platforms, and cloud-based case management. Each transition came with professional responsibility questions. AI presents those questions with far greater urgency and complexity because the technology does not merely automate discrete tasks — it generates content, reasons from facts, and produces outputs that can appear authoritative while being entirely fabricated.

---

JEH Formal Opinion: A Firm-Level Framework

Johnson Evans & Headrick has developed an internal framework for AI use that reflects the evolving guidance from state bar associations and the ABA. The framework addresses three core questions:

1. What AI tools may attorneys use, and for what tasks?
2. What disclosures are required — to clients, courts, and opposing counsel?
3. What supervision obligations apply when AI generates work product?

The firm's position aligns with the emerging consensus: AI tools are permissible and encouraged where they enhance efficiency and accuracy, but attorneys bear final responsibility for every representation made to a client or submitted to a tribunal. The tool does not share the license; the attorney does.

---

Key Themes in Legal AI Ethics

State bar associations, the ABA, and legal ethics scholars have converged on several overlapping themes as they grapple with professional responsibility in the AI era.

Competence

Model Rule of Professional Conduct 1.1 requires that attorneys provide competent representation, including the legal knowledge, skill, thoroughness, and preparation reasonably necessary for the representation. The comment to Rule 1.1 was amended in 2012 to add that competence includes staying current with "the benefits and risks associated with relevant technology."

That comment — sometimes called the "technology competence" amendment — applies directly to AI. An attorney who uses an AI tool without understanding its capabilities and limitations is not competent. This does not mean attorneys must become AI engineers, but it does mean they must understand:

• What the tool is designed to do and not do;
• How the tool generates outputs (probabilistic text prediction, not legal reasoning);
• The known failure modes — particularly hallucination (discussed below); and
• The verification steps required before relying on AI output in any client matter.

Supervision

Model Rule 5.1 (responsibilities of partners and supervisory attorneys) and Rule 5.3 (responsibilities regarding nonlawyer assistance) both apply to AI. When a supervising attorney delegates work to an AI tool, that attorney remains responsible for the output. The duty to supervise does not disappear because the "assistant" is software.

Several bar associations have explicitly analogized AI to nonlawyer staff: the attorney must instruct the tool appropriately, review its work, and take responsibility for corrections.

Confidentiality

Model Rule 1.6 prohibits attorneys from revealing information relating to the representation of a client without informed consent. Entering client-identifying information, case facts, or privileged communications into a public AI tool — one that may use inputs to train future models — raises serious confidentiality concerns.

Attorneys must understand the data practices of any AI tool they use. Enterprise agreements, private deployments, and "private mode" settings matter. Using a consumer-grade AI chatbot with client confidences may constitute a Rule 1.6 violation regardless of whether any actual disclosure occurs.

Candor to the Tribunal

Model Rule 3.3 requires candor toward tribunals and prohibits knowingly making false statements of fact or law. When an AI tool fabricates case citations — which has happened repeatedly — and an attorney submits those citations without verification, Rule 3.3 is directly implicated.

---

Understanding and Integrating AI Into Legal Practice

For attorneys seeking to integrate AI responsibly, the following framework is useful:

Step 1: Understand the tool's design purpose. Is the tool trained on legal data? Is it a general-purpose language model adapted for legal use? Does it search live legal databases or generate from static training data?

Step 2: Understand hallucination. Generative AI models predict the next token in a sequence based on statistical patterns. They do not "know" whether a case exists — they generate text that resembles how a case citation looks. This means plausible-sounding but entirely fabricated case names, docket numbers, and holdings are a known risk of every major language model.

Step 3: Establish verification protocols. Every case citation, every statutory reference, every factual claim in AI-generated work product must be verified against primary sources before being relied upon. AI output is a starting point, not an ending point.

Step 4: Implement confidentiality controls. Use enterprise-grade deployments with appropriate data protection agreements. Do not input client-identifying information or privileged communications into consumer AI platforms.

Step 5: Document your process. Increasingly, courts and clients may ask how work was produced. Maintaining records of AI use and verification steps protects the attorney and demonstrates professional diligence.

---

Competency in Practice: Integrating AI into the Legal Workflow

Competency with AI tools is not binary. It exists on a spectrum that evolves as the technology and the profession's understanding of it evolve. Courts are beginning to establish their own AI disclosure requirements — some requiring attorneys to certify that AI-generated content has been verified, others prohibiting certain AI uses entirely.

The competent AI user in legal practice today is one who:

• Uses AI tools where they add genuine value (research organization, drafting efficiency, summarization of large document sets);
• Does not use AI tools as a substitute for legal judgment;
• Verifies every output before relying on it;
• Communicates transparently with clients about AI use where required or appropriate; and
• Stays current as courts, bar associations, and regulators issue guidance.

---

Case Study: ChatGPT Said This in October 2023

The most-discussed cautionary tale in legal AI ethics is Mata v. Avianca, decided by Judge P. Kevin Castel in the Southern District of New York in 2023. An attorney submitted a brief containing citations to six cases — all fabricated by ChatGPT. When the court attempted to locate the cases, none existed. The attorney, who had relied on ChatGPT without independent verification, was sanctioned.

Judge Castel's order was detailed and unsparing. It concluded that the attorney had submitted false statements to the court and failed to correct them even after opposing counsel raised concerns. The sanctions included monetary penalties and a requirement that the court order be sent to the clients and the state bar.

Mata v. Avianca is not an isolated incident. Since 2023, courts around the country have encountered fabricated AI citations in filed documents. Some attorneys have faced bar complaints. Others have received warnings. The consistent message from the bench is: you are responsible for what you file. AI is not an excuse.

---

Case Study: The Difference Between AI That Assists and AI That Replaces

There is an important distinction — one that plays out in practice every day — between using AI as a tool that assists attorney judgment and treating AI as a replacement for it.

AI that assists: An attorney uses an AI research tool to identify potentially relevant cases in a jurisdiction she is unfamiliar with. She reviews each result, reads the full opinions, verifies that the cases stand for the propositions she intends to cite, and incorporates her own analysis. AI has saved hours of research time and produced a better work product.

AI that replaces: An attorney pastes a client's fact pattern into ChatGPT, receives a "draft brief," and files it with minimal review. She assumes the citations are real because they look real. They are not.

The professional responsibility framework does not prohibit the first and condemns the second. The technology is the same; the professional judgment is not.

---

Case Study: Clinical AI Integration in Legal Practice

Several major law firms have piloted "clinical" AI integration programs — structured environments in which AI tools are deployed for specific, supervised tasks with established quality-control protocols. Early results from these programs demonstrate:

• Significant reduction in time required for first-draft contract review;
• Improved consistency in identifying standard risk clauses;
• No significant reduction in attorney satisfaction with the quality of final work product when attorneys properly review AI outputs.

What these programs also consistently find: the attorneys who get the most value from AI tools are those who invest time in understanding how those tools work and in developing reliable verification habits. The technology rewards engaged users and punishes passive ones.

---

How Do You Become Competent in AI?

Several practical pathways exist for attorneys seeking to develop AI competence:

1. CLE programming. State bar associations have responded to attorney demand with AI-focused CLE, including courses on the technology itself, professional responsibility implications, and jurisdiction-specific guidance. Many states now offer CLE credit specifically for technology competence programming.

2. Hands-on use. Reading about AI tools is not the same as using them. Attorneys should experiment with general-purpose and legal-specific platforms in low-stakes settings — drafting internal memos, summarizing documents for their own use — before deploying AI in live client matters.

3. Bar association formal opinions. The ABA and multiple state bars have issued or are drafting formal opinions on AI. These opinions are the profession's closest analog to a regulatory framework. Attorneys should read and monitor guidance from their state bar.

4. Peer networks. Law firm practice groups, bar sections, and listservs dedicated to legal technology are valuable resources. Attorneys who are actively grappling with these issues share practical wisdom that no formal opinion can fully capture.

5. Vendor diligence. For attorneys considering enterprise AI tools, diligence on the vendor — including data privacy practices, accuracy benchmarks, and incident history — is part of competent adoption.

---

Copyright and AI: The Emerging Legal Landscape

The intersection of AI and copyright law is among the most contested areas of intellectual property today. Two distinct questions drive most of the litigation and commentary:

Question 1: Can AI-generated content be copyrighted?

The Copyright Office has consistently held that human authorship is a prerequisite for copyright protection. A work generated entirely by AI — without creative input from a human author — is not copyrightable. The Copyright Office's 2023 guidance on AI-generated works confirmed this position: copyright protects human expression, not machine output.

This does not mean that works incorporating AI assistance are unprotectable. Where a human author makes sufficiently creative selections, arrangements, or modifications to AI-generated content, the human-authored elements may qualify for protection. The line between "AI-assisted" and "AI-generated" is still being drawn.

Question 2: Does training AI on copyrighted works infringe those works?

This question is the subject of a wave of litigation against major AI developers. Authors, artists, musicians, and news organizations have filed suit arguing that training large language models and image generators on their copyrighted works — without license or compensation — constitutes infringement.

The defendants generally argue:

• Training is a transformative use protected by the fair use doctrine;
• The models do not store or reproduce the training data; and
• The output of the model is not substantially similar to any specific training work.

The plaintiffs argue:

• Training copies the works without license;
• Fair use does not extend to commercial reproduction at scale; and
• The outputs often reproduce distinctive elements of the training data.

Courts have yet to reach consistent conclusions. Some cases have been dismissed; others are proceeding to discovery. The Supreme Court has not yet addressed the question directly.

---

What Copyright Uncertainty Means for Attorneys and Clients

For law firms and clients who create, commission, or rely on AI-generated content, the copyright uncertainty has practical implications:

Ownership questions. If AI-generated content cannot be copyrighted, works produced using AI tools may fall into the public domain immediately. Clients who invest significant resources in AI-generated content should understand that protection may be limited or unavailable.

Licensing risk. Clients who build products or services using AI models that were trained on copyrighted data may face downstream infringement exposure if courts ultimately hold that unlicensed training constitutes infringement.

Disclosure in registration. The Copyright Office requires applicants to disclose AI-generated content in registration applications. Failure to disclose may result in invalidation of the registration. Attorneys advising clients on copyright strategy must account for this.

Chain-of-title diligence. In transactions involving AI-generated content — licensing, acquisition, investment — proper diligence now includes understanding what AI tools were used, what data those tools were trained on, and what copyright representations the counterparty can actually make.

---

A Practical Ethical Checklist for AI-Using Attorneys

Before deploying an AI tool in client matters, attorneys should work through the following:

• Competence: Do I understand how this tool works and its known limitations?
• Confidentiality: Does this tool have appropriate data protection for client information?
• Supervision: Have I reviewed and verified the AI's output — every citation, every factual claim?
• Candor: Am I prepared to certify the accuracy of this submission?
• Disclosure: Does my engagement agreement, the court's standing orders, or applicable bar guidance require disclosure of AI use?
• Documentation: Have I retained records of my AI use and verification process?

---

Conclusion: Embracing AI Without Abandoning Professional Judgment

The legal profession has always adapted to new tools. AI is a more powerful and more unpredictable tool than most. It can make attorneys dramatically more efficient; it can also expose them to sanctions, bar complaints, and malpractice claims if used carelessly.

The ethical path forward is not technophobia. Refusing to understand or engage with AI tools is itself a competence failure as the profession increasingly expects attorneys to deploy them responsibly. The path forward is engaged, skeptical, verified use — AI as an assistant that augments professional judgment, never as a substitute for it.

The attorneys who thrive in the AI era will be those who understand the technology, apply it with discipline, stay current with evolving guidance, and never forget that their name — not the model's — is on the filing.

---

This article is based on a presentation delivered to the American Inns of Court by Michael A. Johnson of Johnson Evans & Headrick, P.C. It is for informational purposes only and does not constitute legal advice. For guidance on AI use in your practice or AI-related intellectual property matters, contact our team.