Lawyers continue to grapple with AI ethical issues

Despite the widening acceptance of generative artificial intelligence tools within the legal industry, uncertainty remains about the ethical implications of the still-evolving technology for lawyers and their clients.

And it remains unclear whether the canons of professional conduct, the bulk of which were developed before GenAI’s dramatic emergence, remain adequate to guide lawyers’ conduct in a future where AI use seems only more likely to proliferate.

Nearly 7 in 10 legal professionals report using some form of generative AI, ranging from general-purpose large-language (LLM) platforms like OpenAI, ChatGPT, and Google Gemini to legal industry-specific tools like CoCounsel, Everlaw, Harvey AI and Lexis+ AI, according to a survey of 1,300 firms in September and October 2025.

The same survey, conducted by the legal technology company 8am, had found GenAI use at 27% among legal professionals in 2024 and 31% in 2024.

Hallucinations, deepfakes and other risks

Ryan Groff, a lecturer at New England Law School in Boston, said existing ethical standards have been sufficient to let courts punish attorneys for allowing the best-known deficiency of legal AI – “hallucinated” or make up cases, complete with imaginary citations – into their legal filings. Disciplinary cases involving AI-generated hallucinated cases first began appearing in court filings in 2023 and 2024.

“While AI hallucinations are the product of newer technology, they don’t require new responsibility from legal professionals,” Groff said. “We’ve had rules on the books for years requiring lawyers to oversee the assistance they receive. Decisions requiring legal competence were always to be made by licensed legal professionals.”

Jessiah S. Hulle, a litigator with Gentry Locke in Roanoke, Virginia, said that a general rule of ethics regarding AI use is that attorneys must be aware of the technology’s capabilities and limitations, whether they are using it themselves or not.

“If you are using generative AI, you should know about the risks involved, to avoid violating ethical rules, court orders, or procedural rules,” he said. “But even if you are not using AI, you should pay attention to both its potential benefits and potential issues with others using it. You don’t want to be caught off guard by, say, an expert relying on AI for a declaration, or a party producing deepfake evidence in discovery, or a client exposing confidential information to an open source chatbot.”

Model rules and state requirements

In 2024, the ABA’s Standing Committee on Ethics and Professional Responsibility issued Formal Opinion 512. While not binding itself, the opinion held that core Model Rules contain provisions that also apply to AI use by lawyers in several aspects of their practices: Rule 1.1 (Lawyer Competence), Rule 1.6 (Lawyer-Client Confidentiality), Rule 1.4 (Communications with Clients), Rule 1.5 (Fees and Billing), Rules 5.1 and 5.1 (Lawyers’ Supervisory Duties), and Rule 3,3 (Candor toward the Tribunal).

In general, the rules require lawyers to keep informed of the latest developments on the legal AI landscape, to be responsible for AI-generated content they adopt as their own work or that of those they supervise, to be forthright with clients about AI use in their cases, and to protect confidential client information from disclosure due to unsecure AI tools.

Most state bar associations or attorney supervisory agencies like court administrators have since layered other AI-related requirements onto the Model Rules framework.

For example, state bar regulators in New Jersey require legal AI instruction to be part of the state’s mandatory continuing legal education regimen. Other states are poised to follow suit.

In May, Florida’s Supreme Court amended the state’s rules of practice to require the filers of all documents before its courts to verify that “the legal authorities identified exist and are accurately cited.”  The judges noted that the use of generative artificial intelligence tools for document preparation by both lawyers and pro se litigants was becoming more common and that such aids “can be helpful.” But the court cautioned that AI “also can generate content that appears plausible but is in fact inaccurate, including fabricated or ‘hallucinated’ authorities.”

In New York, a new rule scheduled to go into effect in mid-August specifically says all lawyers appearing before courts in the state must understand AI’s capabilities and limitations and ensure that all submissions contain no fictitious citations or other AI-generated errors.

A ‘competency problem’

Nicole Black, a Rochester, New York, lawyer who authored the AI-use survey results as Principal Legal Insight Strategist at 8am, said the verification of court submissions being required in New York was “lawyering 101.” In her view, existing ethics guidelines are more than sufficient to punish AI offenders.

Attorneys using bad AI-generated information don’t have a “technology problem,” Black said, but a “competency problem” because they do not understand their basic obligations as attorneys.

“Just because courts are faced with a tsunami of verifiably inaccurate filings doesn’t mean we need new rules,” she said. “Instead of passing redundant regulations, we should continue to educate the profession about the risks of AI and continue to penalize practitioners for negligent lawyering.”

Barbara M. Seymour, a professional responsibility and legal ethics specialist at Clawson & Staubes in Columbia, South Carolina, agreed that client disclosure, confidentiality, and fee issues involving AI tools are “easily” addressed by existing ethical canons. In general, more AI-specific rules are not needed to guide attorney conduct, she said.

In Seymour’s view, bar associations can better advance the continuing education of attorneys about the pitfalls and limitations of AI use by providing “practical guidance and training” for practitioners.

A toolkit on generative AI use from the Georgia State Bar is particularly helpful, Seymour opined. She also noted a formal opinion from the North Carolina State Bar as containing useful guidance on AI use.

More recently, AI use had led to a newer issue: Clients who have researched their own cases using generative AI. Those clients, said Seymour, come into conferences armed with their own legal analysis, strategic demands and “unrealistic expectations.”

“This is the equivalent of the doctors’ WebMD problems,” she said, where patients come to doctors with preconceptions and often misconceptions of what’s wrong with them based on the internet research.

Confidentiality concerns

Grace H. Wynn, an attorney at HWG in Raleigh, North Carolina, and lecturing fellow at Duke University School of Law, said the preservation of confidentiality in the use of legal AI tools is an area that deserves more ethical guidance.

“The confidentiality rules were not written with AI in mind, and the ethics opinions on this subject fail to fully address that gap in guidance,” she said. “Given the massive consequences of privilege waivers and confidentiality breaches, more definitive guidance is needed.”

But whether practitioners realize it or not, lawyers’ intellectual makeup and their penchant for exhaustively researching questions can help protect them from AI-generated pitfalls, Groff suggested.

“The best legal professionals tend to be curious,” he said. “They will turn a question around and around, making sure they consider it from different angles. Pre-GenAI, they knew legal research required many, many searches; not just one query/filter combination. When they use AI with that same degree of curiosity, I think they tend not to misuse it.”