In my practice, I receive lots of inquiries surrounding the use of AI by attorneys, including issues of confidentiality, competency, passing along AI costs to clients, disclosure to clients, truth in advertising, etc. An excellent ethics opinion on the topic is 2024 FEO 1. Recently, however, there has been a shift in those inquiries, from the use of AI by attorneys to the use of AI by clients. Attorneys have concerns about clients using AI to draft legal documents and wanting the attorney to “just review this.” In addition, clients who have received drafts of legal documents from the attorney for review have been using AI to edit and comment on these documents.
The clients’ use of AI raises several issues under the ethics rules. First, an attorney cannot merely review an AI-generated document from a client or prospective client and provide advice. Without having the appropriate conversation with the client, gathering sufficient information, engaging in appropriate due diligence, and having a thorough discussion with the client about the client’s goals, the lawyer risks a violation of Rule 1.1: Competency and of Rule 1.4: Communication. Rule 1.1 requires that an attorney have “the legal knowledge, skill, thoroughness, and preparation reasonably necessary for the representation.” An attorney who merely reviews an AI-generated legal document has not engaged in the appropriate due diligence to provide thorough and competent representation. Under Rule 1.4(a), an attorney has a duty to “reasonably consult with the client about the means by which the client’s objectives are to be accomplished[.]” Without having a discussion with the client about the client’s objectives, an attorney cannot fulfill his or her duties to the client under this Rule.
Attorneys are also seeing clients uploading legal documents the attorney drafted to various AI platforms. The clients then come back with extensive edits and pages and pages of comments driven by AI. Attorneys have expressed concern about the additional time it takes to review these edits and comments, and then to explain to the client why AI was wrong or why a particular edit would not be in the client’s best interest. Further, the client does not understand that this additional review can make their legal fees (if hourly billing) rise dramatically. If the lawyer is billing on a flat fee basis, this kind of additional and unanticipated work means the lawyer will likely lose money on the representation. I have advised lawyers to be sure that the client understands on the front end, or before agreeing to review the AI-generated end product, what using AI means, including that it could mean a significant increase in fees for the time it takes to review the AI-generated materials. An attorney should try to manage the client’s expectations, especially if fees likely will be significantly increased unexpectedly.
More concerning to me, however, is that the client likely does not realize that uploading drafts of documents containing confidential information could result in an unintended waiver of the attorney-client privilege. Given this AI upload trend with clients, I am advising lawyers to include in their engagement agreement, or in some written format at the beginning, a warning to clients not to upload their legal drafts to Open AI or other AI systems. The lawyer does not have control over those systems and can’t reasonably evaluate whether each platform that a client may use is secure, so the better practice is to advise clients not to use those platforms. That advice is not only prudent because it helps protect attorney-client privileged information, but it may also have the practical benefit of limiting voluminous AI output presented by clients.
Finally, we are seeing a significant increase in AI-generated State Bar complaints, as well as complaints with other licensing boards. These kinds of AI-generated complaints, given their length and the pace at which they can be created and submitted, are causing issues for both professional licensees and the Boards regulating their conduct. This trend does not seem to be slowing down. If you are seeing other kinds of client AI use in your practice, feel free to let me know.
