Whether we like it or not, as it currently stands, artificial intelligence (AI) is shaping the future of the legal profession. More than changing how lawyers review documents and draft pleadings, artificial intelligence is also changing how clients act. For example, before reaching out to an attorney, a potential client may put the facts of their matter into a generative AI platform and ask the chatbot if they have a chance of winning if they pursue a lawsuit. Once a client has an attorney retained, they may send things to their lawyer that have been drafted or summarized by AI. As lawyers, we need to be aware of the risks that this behavior creates for our clients and advise them accordingly. It is imperative that lawyers explain to clients that the use of AI could jeopardize confidentiality.1
WHAT IS GENERATIVE AI?
AI-powered tools offer capabilities that mirror key legal skills, including analytical and research abilities, attention to detail, persuasive writing, organization, and interpersonal communication. Generative AI is a subset of artificial intelligence capable of creating new content (images, text, videos, audio clips, etc.) in response to prompts, based on a large reference database of examples. Regular old artificial intelligence has been around for more than 30 years. It is used in all kinds of things from spelling and grammar check software to smartphone photography, email spam filtering, Google Maps, and your Facebook timeline. Those who say they do not use AI probably mean that they do not use generative AI.
Generative AI is what most people think of when they think of AI. It came into the public’s consciousness with the release of ChatGPT to the public in 2022. It generally refers to an artificial intelligence program that recognizes patterns and is designed to generate content rather than provide factual answers. Many articles have been written about the impact of generative AI on the practice of law.2 We all know that generative AI is having a significant impact on day-to-day legal work. One risk you may be less aware of, however, is the impact these generative AI programs, and our clients’ use of them, will have on client relationships.
CONFIDENTIALITY
Lawyers have an affirmative duty under MRPC 1.6 to protect their clients’ confidences and secrets. This requires lawyers to protect clients’ confidences and secrets when inputting client information into any type of program, including generative AI programs. Under MRPC 1.6, lawyers must maintain client confidences and cannot disclose information without informed consent unless an exception applies. AI tools pose new challenges in ensuring confidentiality, making it essential for lawyers to carefully evaluate how these tools handle data. For example, conversations you have with an AI program may be reviewed by a human if the program detects a safety risk. If content is flagged as high risk, it may be reviewed by a real person. Further, depending on the AI program and the settings on your account, your conversations may be reviewed at any point by human trainers to assist them in improving the AI model. Further, information input into an AI program may be used by the model in a conversation with another user. The risks to confidentiality when using these tools are substantial.3 The focus of this article, however, is on client behavior.
CLIENT USE AND RISK
One major risk generative AI poses is the fact that clients may be unaware that the information they enter into these programs may not be confidential. Information entered into AI prompts is often incorporated into the learning model or reviewed for safety or abuse detection, risking client confidentiality.
Sam Altman, the CEO of OpenAI, recently said in a podcast interview that “if you go talk to ChatGPT about your most sensitive stuff and then there’s like a lawsuit or whatever, we could be required to produce that.”4 Many users (if not most) still assume that anything they type into a generative AI prompt is private. Clients may believe that their prompts are temporary and private, but it is very possible that everything input into a generative AI tool is being stored, reviewed, and preserved. And, if requested, inputs may need to be produced in discovery. For litigators, the danger of this discoverability is apparent. For clients, it is less obvious.
This issue has already been through the courts in at least one instance. In United States of America v. Bradley Heppner,5 Heppner used Anthropic’s AI tool Claude, a generative AI tool, to input his defense strategy, seeking comments on his strategy and potential arguments, and he asked what the other side might argue. Then, he gave all the information he received from Claude to his attorney. Heppner was criminally charged with fraud-related activities, his electronic devices were seized, and the AI communications were obtained by the government. Heppner’s lawyers argued that Heppner utilized Claude for legal advice, which should be protected under attorney-client privilege or the work product doctrine. The judge ruled that the written exchanges between Heppner and Claude were not protected from disclosure to the government by the work product doctrine or attorney-client privilege.6 Importantly, Heppner did not tell his attorney in advance that he was planning to use AI.
Heppner is far from the only client inputting information about his legal matter into a generative AI platform. When a client uses a public tool, like ChatGPT or Copilot, without opting for enterprise-level privacy protection, inputting sensitive information might be considered disclosure to a third party. If the provider of the tool maintains the ability to retain, review, or use the input information, courts may find that confidentiality has not been preserved. Thus, it is imperative in today’s age for lawyers to educate their clients about the risks and benefits of using these tools.7
Further, while Heppner is a criminal case, the same applies in the civil context. Forensic evaluations of electronic devices occasionally occur in the context of civil litigation. Now that the court in Heppner has affirmatively stated that the information entered into an AI chatbot may not be protected, subpoenas may follow. A further risk posed by using these AI tools is that clients’ accounts may be vulnerable to hacking by third parties.
The risks that are posed by AI chatbots providing legal advice go beyond the risk to confidentiality. States are starting to take action in light of these risks. For example, New York is currently considering a bill that would ban these AI platforms from offering legal or medical advice.8 However, the wheels of bureaucracy turn slowly, and until regulation catches up with reality, lawyers should be aware of these risks and make sure clients are aware of them as well.
CLIENT RELATIONS
Further, if a client is coming to you, it is because they think their cause is good. Now, with generative AI tools, which are programmed to agree with you and tell you what you want to hear, clients will believe even more strongly in the rightness of their cause. We all need to prepare for the impact this shift will have on client relationships. Clients will likely increasingly rely on generative AI programs for help with legal advice and strategy. Given the fact that these tools tell people what they want to hear, give false information, and even completely hallucinate, this will create a big shift in the lawyer-client relationship.
If lawyers are not careful and purposeful in their approach, they run the risk of eroding the trust that clients place in them, and trust has always been the foundation upon which the attorney-client relationship stands.
CONCLUSION
Awareness is the first step in preparing your practice for the changes AI is bringing about. Lawyers should be aware of the risk their clients put themselves at by using generative AI programs and warn their clients of any risks to the confidentiality of the information they input into these programs. Further, in the age of AI, lawyers must do what AI can’t: form real, personal connections with clients based on trust. Through preparation, knowledge, and empathy, lawyers can earn a client’s trust and stay one step ahead of this emerging technology.