Mid-Atlantic Health Law TOPICS

AI is everywhere, and in many cases, it is an advantage actively to incorporate AI into daily life. But, by choosing to engage with large language models, such as ChatGPT, Claude, Gemini and others, you could be giving up the significant legal protection of attorney-client privilege.

Attorney-client privilege protects certain communications between a client and the client’s attorney that were made for the purpose of obtaining legal advice from being disclosed to a third-party in a legal proceeding. This privilege allows clients to be candid and truthful with their legal counsel, without concern that the communication could end up in the hands of an adverse party.

Sharing Legal Advice with AI 

Recently, courts in the United States have been considering whether that privilege continues to exist once clients share communications they received from their attorney with a publicly available AI platform. At least one judge in a district court in New York determined that by sharing their lawyer’s communication with a public AI, the client had compromised the confidentiality of the communication and the privilege no longer applied to those communications. 

The same judge rejected arguments that the public AI was an agent of the legal counsel, which in some cases can preserve the privilege. The judge rejected the “agent” arguments  because (1) the AI is not human and (2) the attorney did not direct the client to share the communication with the AI. 

The result is that where communications between a client and a lawyer are shared with a public AI, they may be discoverable and required to be disclosed to an adverse party in a legal case. 

This also has implications for the various AI notetakers that are built into popular video calling platforms. Often automatically launched when the call begins, the AI notetakers could be gathering information during a call between an attorney and his or her client, compromising the privilege of the conversation.

Representing Yourself

In contrast, a judge in the Eastern District of Michigan recently held that where an individual was representing herself in an employment dispute, her interactions with a public AI were not discoverable under the Work Product Doctrine. The Work Product Doctrine protects the writings of legal counsel that are created in preparation for or in anticipation of litigation. 

Here, where the individual was not a lawyer but was representing herself, the judge found that the Work Product Doctrine would apply. Additionally, the judge found that the public AI was a “tool” not a person and, therefore, the individual had not disclosed the information to an adversary (or disclosed it in such a way that it would be likely to be discovered by an adversary) thereby voiding the Work Product Doctrine.

Conclusion

As AI tools become woven into everyday life, it is important to remember that convenience can come with hidden costs. You should treat public-facing AI platforms as you would any other third party: assume what you type may not remain confidential; and don’t paste in sensitive facts, documents, or communications with counsel, unless your attorney has specifically approved that approach. 

You should check to ensure that your video calls are not being monitored by AI notetakers that may not be secure, and if one launches automatically, you should remove it from the call before receiving advice from legal counsel.

Used thoughtfully, AI can help you organize ideas and understand general concepts—but when legal rights, strategy and privilege are on the line, the safest path is to slow down, protect confidentiality, and get guidance from a licensed lawyer before you share or rely on anything generated by AI.

Darci M. Smith
410-576-4153 • dsmith@gfrlaw.com