It has already become common that a person facing a legal issue will first attempt to consult some Artificial Intelligence (AI) portal better understand the issue. This involves the person entering prompts that explains the issue and then engaging in a conversation with AI where additional information is given and questions are asked by both the person and AI to each other.
This article could be about the fact that at the end of the day AI is simply another computer program, albeit self-learning, and thus is not immune from something known as “GIGO”, meaning garbage-in garbage-out. Persons consulting AI for legal advice may not know the facts that are relevant, or may downplay certain facts, so that when AI computes all this it comes up with a flawed response. But this article is not about that.
This article could also be about how persons who do this then take their AI output to their attorney for review, and the attorney will end up spending more time ferreting out the real facts and figuring out why the AI output is incorrect than if the attorney had just started from scratch in the first place. But this article is not about that either.
What this article is about is the conversation that a person has with AI, including the facts that they provide to AI. This is where the danger lies.
All U.S. jurisdictions have attorney-client privilege. Depending on the jurisdiction, there are also spousal privileges (communications between a married couple), accountant privileges, and the like. These privileges create a important protection for qualifying communications such that an adverse litigant may not obtain discovery of these communications or use them in court.
AI communications are not protected by any of these privileges.
Let’s focus on attorney-client privilege since that tends to be the privilege most often asserted in litigation. For the privilege to apply, the communication must generally satisfy two elements. First, the communication must be to a licensed attorney. Second, the communication must have been made in anticipating of receiving legal advice. While there are exceptions, such as when the communication is for the purpose of committing a crime or fraud, in general this privilege will protect communications when those two elements are met.
AI communications fail with both elements. First, AI is not a licensed attorney, at least not yet. Second, deep within the legal disclaimers for AI portals there is something to the effect that AI is not giving or receiving legal advice or counsel and it may not be relied upon for legal purposes (if there is no such disclaimer, then feel free to bring your class action against the AI portal if the advice turns out wrong). This causes the second element to fail as well, even if it sure seemed like dispensing legal advice was what AI was doing. And, no, it does not matter that you later shared the conversation transcript or the conclusion or other output with a licensed attorney, that does not make it retroactively privileged. Doesn’t work that way.
The bottom line is that communications or conversations or interactions with AI are not protected by privilege. This means that they are fair game for discovery by an adverse party in litigation. Consider that an adverse party might serve the following requests upon you.
Interrogatory No. 17: Identify any and all artificial intelligence services, however described, that you have utilized in the last twelve (12) months.
Interrogatory No. 18: Identify all communications that you have had with any and all artificial intelligence service regarding the subject matter of this lawsuit, including chats or the use of AI agents, and state the date(s) upon which communications occurred and the subject matter of those communications.
Document Request No. 29: Produce all transcripts of conversations, including chat logs and similar documents, and also including any conclusions reached, which you have had with any artificial intelligence service relating to the subject matter of this lawsuit.
If you have had such conversations and have saved or printed out those conversations, then you are required to identify and disclose them. But, you say, what if I just fudge the truth (lie) and say I never had such a conversation or never saved the transcript or output?
Not a good idea. First, if you have already even just told your attorney about these conversations, or maybe even printed out and provided them to your attorney, then she may be required as an officer of the court to disclose them to the adverse party in the discovery responses (or not, depending upon the circumstances). Second, it is the felony of perjury.
Now, if you are thinking that the the adverse party will never discover your communications with AI, consider that the case goes forward and you lose a judgment. As part of judgment enforcement efforts, the adverse party runs a credit report and sees that you have a MasterCard issued from a certain bank. The adverse party then issues a subpoena to that bank and obtains your credit card statements for the last four years. Those credit card statements show that you were billed for using an AI service. The adverse party now services a subpoena for all records relating to your use of the AI service and, bingo, you are caught lying under oath.
Because folks think that they’ll never have to disclose their AI communications, they are painfully indiscrete. Debtors will start their AI chats with such questions as, “How do I keep my crypto account from being found by creditors?” Yeah, if a creditor gets a hold of that, it’s going to go down with the judge like a lead salami. Goodbye bankruptcy discharge too.
By now you should get the point: If you are in litigation, or might soon be in litigation, then you don’t want to be stupid in how you use AI in relation to something you want to know about. Please do not attempt to do your own research and then provide it to your attorney in the hope that it will save a few bucks in legal research. First, it usually will not and, second, it opens up a big can of worms.
Having said all this, it should be noted that this is one of those emerging areas in litigation where we as attorneys can’t really say in advance how all of this is going to finally shake out. There might be situations where it could be successfully argued that a person was using AI to simply organize their thoughts as a prelude to meeting with their attorney and seeking legal advice which would be privileged.
But like all emerging areas of the law, you don’t want to be the test case.