When AI Gets Between You and Your Lawyer
Artificial Intelligence (AI) has become part of everyday life. Tools such as ChatGPT or Claude can draft analyses and explain information with remarkable speed. This can be useful, but there are perils.
Much has been written about the consequences when lawyers use AI and simply accept what it generates. But another phenomenon has drawn less attention: the growing number of clients who use AI to communicate with their lawyers.
The impulse is understandable. Clients in emotional turmoil who face a high-stakes situation may want to be sure they explain everything, organize their thoughts clearly, and clearly articulate what they need to say.
While it can be helpful to use AI, it’s important to understand the risks of doing so in a lawyer-client relationship.
Confidentiality
A lawyer's advice to a client is generally protected and privileged. Feeding legal communications, advice, or any sensitive facts into an AI system may introduce another party into that chain of communication.
Clients should never put confidential or privileged information into an open-source AI chatbot. Doing so may create serious confidentiality concerns and, in some cases, raise questions about whether privilege has been compromised or waived. Removing names and other identifying information can reduce the risk, but it does not necessarily eliminate it. A person may still be identifiable from the surrounding facts, particularly where the circumstances are unusual or distinctive.
When in doubt about whether something should be put into AI, do not do it.
AI Can Complicate Matters
One of the most important limitations of generative AI is also one of its greatest strengths: It produces fluent, persuasive responses that can sound more reliable and compelling than they are.
AI can fill gaps in information with assumptions. It can turn a possibility into an apparent fact. It can identify a legal principle that sounds relevant but does not apply to the circumstances at hand. It can take a client's description of events, combine it with its own assumptions, and produce a confident analysis that simply isn't correct.
The result can be problematic because legal advice depends on judgment and context, not just facts and arguments.
This can make the lawyer-client relationship less efficient.
Consider an ordinary exchange where a lawyer sends a client a draft document. The client has questions such as, “What does this provision mean?” or “Can this section be reworded to emphasize X instead of Y?”
Increasingly however, the client will feed the draft into an AI tool and pass along a lengthy analysis instead of the question at the core of their concern. It may feature a summary of the document, proposed revisions, alternative legal theories, background information, a list of potential issues, and a series of follow-up questions.
This can be unhelpful, especially if much of the AI-generated response incorporates incorrect assumptions or includes legal concepts that are irrelevant to the situation in question.
The lawyer then faces new tasks: to decipher the AI-generated response, try to determine the client’s primary concerns, identify which assumptions are wrong, and respond to each of the issues raised by the AI response. For counsel, this adds to the time spent on the file. For the client, it adds to costs.
A Good Way For Clients to Use AI
There are situations where AI can be remarkably useful.
For example, in instructing a lawyer for preparation of a power of attorney for personal care, a client may struggle to articulate their very specific wishes about what they would want to see happen if they were to become incapable of making decisions. Using AI to organize or express those thoughts can help communicate those intentions to their lawyer. The lawyer can apply legal expertise and judgment and translate the client’s instructions into appropriate legal language.
The same can be true in litigation: AI can be useful as a private tool to help a client organize their thoughts, identify concerns, or propose topics to discuss with their lawyer. The problem arises when the client treats the AI analysis as authoritative or sends their lawyer an AI-generated legal analysis instead of explaining their underlying concern.
The important distinction is to use AI to assist in a client's thinking, not as a substitute for the lawyer's judgment; nor should it become an intermediary through which the client attempts to direct the legal work.
A lawyer's role is not just to take instructions and put them into polished language. It is to understand the client's circumstances and objectives, identify the legal issues that matter, exercise professional judgment and advise accordingly.
The best legal instructions are not necessarily the most polished ones. They accurately communicate what the client knows, what the client wants, what the client is concerned about, and what the client does not understand.
The goal is not to keep AI out of the lawyer-client relationship, but to use it discreetly. Inserting the views of another “being,” one that is not fully informed, can hinder far more than it can help.