Lawyers catering to affluent clients are increasingly encountering a new challenge: their clients are using artificial intelligence chatbots like ChatGPT and Claude to analyze their estate plans and seek tax strategy suggestions. This trend, while sometimes leading to more informed client questions, is also creating significant headaches and potential legal risks for both the clients and their legal counsel. Many high-net-worth individuals, even those who can readily afford premium legal services, are turning to AI for its perceived convenience and cost savings, a phenomenon that has prompted lawyers to spend more time defending their advice and protecting clients from potentially erroneous information.
Tasha Dickinson, a partner at Day Pitney, said she receives calls every week from clients inquiring about legal advice they’ve obtained from AI chatbots such as ChatGPT and Claude. While some clients are upfront about their AI consultations, Dickinson noted that she can often discern their AI-driven queries by the nature of their questions. She recounted an instance where a high-net-worth client in Florida proposed establishing a community property trust, a strategy suggested by AI to reduce taxes for his heirs, she said.
Dickinson had to gently inform the client that such a trust is specifically for married couples, and his wife had recently passed away. “Well, you do understand that a community property trust is between husband and wife, right?” Dickinson reportedly asked, to which there was silence on the phone. The client’s surprised silence underscored the disconnect between AI’s generalized suggestions and the specific, nuanced realities of an individual’s legal situation, with Dickinson adding, “They’re like, ‘Oh, well, AI thought it was a good strategy.’
Well, like, in the universe, maybe it’s a good strategy, but it’s not a good strategy for you.”
Robert Strauss, a partner at Weinstock Manion, has observed a similar pattern. Several clients have begun uploading their trust documents into AI systems and returning with lists of questions and proposed amendments, forcing Strauss and his colleagues to justify their existing strategies and explain why AI-generated recommendations may not be suitable for their clients’ unique circumstances.
While Strauss acknowledged that these AI-prompted questions can sometimes lead to more informed discussions, he noted that it significantly increases the time spent on matters that have, thus far, yielded no practical or workable suggestions. “The questions are fine, but it results in spending more time on the matter than we would ordinarily spend,” he said. “We end up spending two, three, four hours of time dealing with stuff that so far has amounted to nothing. I have not actually received a single workable suggestion from that process.”
This process, he fears, can erode the client’s trust in their legal advisor.
Beyond the time investment and potential for client distrust, a more profound concern for lawyers like Strauss is the issue of data privacy. Clients are sharing highly sensitive personal and financial information with large-language models, raising significant privacy concerns and creating potential legal vulnerabilities. Strauss said his firm is actively revising its client contracts to include explicit warnings that the use of AI chatbots for legal research or advice could potentially void attorney-client privilege.
This concern is not theoretical; in February, a federal judge reportedly ruled that a criminal defendant’s discussions with Claude regarding his legal defense strategy were not protected by attorney-client privilege, highlighting the precariousness of confidentiality when engaging with AI.
Dickinson emphasized that her primary apprehension regarding AI is not its occasional inaccuracies, which she can correct, nor the fact that clients are double-checking her work, as she has strong confidence in her own expertise. Instead, her significant concern lies in the potential for clients to inadvertently waive attorney-client privilege by inputting confidential documents and queries into AI systems. “What I am concerned about is that when people put documents and do these searches into AI, they’re waiving the attorney-client privilege, and that is a huge issue,” she said. This, she believes, represents a substantial risk that could have far-reaching implications for legal protections.
Dan Griffith, director of wealth strategy at Huntington Bank, echoed these concerns, warning that seeking AI-generated advice on complex matters such as asset protection through prenuptial agreements or tax-efficient business sales could be used as evidence against individuals in legal proceedings. While affluent clients can typically access and afford top-tier legal counsel, the allure of AI’s convenience is undeniable, a sentiment shared by many across all socioeconomic strata, according to Griffith. The perceived cost savings associated with AI also remain a significant draw for clients, as Dickinson pointed out that no one enjoys paying for professional services.
Many of the successful entrepreneurs among Dickinson’s clientele possess a strong drive for knowledge and a high degree of self-confidence. This can sometimes lead them to overestimate their understanding of complex legal and financial matters. Dickinson suggests that using AI tools can foster a “false sense of knowledge,” where clients believe they grasp a subject more thoroughly than they actually do. This phenomenon is not entirely novel; clients have historically brought suggestions from friends, articles, or even informal discussions to their lawyers. Dickinson characterized the current AI trend as “a more evolved form of cocktail party talk.”
While many legal professionals also utilize AI in their work and personal lives, leading to some high-profile errors like briefs containing fabricated citations, clients often lack the specific expertise in both AI and the law to craft effective prompts, lawyers said. Ed Renn of Withers shared an example of a client who, based on ChatGPT’s advice, intended to transfer an unlimited amount of assets to his spouse. However, the client failed to disclose that his wife was foreign-born. This crucial detail would have prevented him from utilizing the unlimited marital deduction without the establishment of a specific type of trust, according to Renn. “If you don’t know quite what you’re doing, it’s garbage in, garbage out,” Renn reportedly summarized.
Renn further noted that AI tools tend to struggle with more intricate subjects, such as international taxation, and may not always be up-to-date with the latest legislation or guidance from the Internal Revenue Service. Griffith highlighted that the process of transferring wealth to beneficiaries is inherently complex, involving intricate discussions that current AI models are ill-equipped to handle. He illustrated this with the challenge of dividing assets between children from a previous marriage and a current spouse.
Griffith explained that a simple ‘yes’ or ‘no’ answer from an AI regarding a child’s access to inherited funds is insufficient. A proper legal consultation requires a deeper exploration of family dynamics and personal circumstances, a level of qualitative analysis that AI currently cannot provide. He observed that AI often prioritizes finding a solution, sometimes defaulting to a ‘yes,’ without adequately probing the core of the client’s underlying question or concerns.
