I recently wrote about the ruling of a federal district court judge in U.S. v. Heppner. The gist of my article was that the ruling was a warning that materials to and from publicly facing GenAI tools were discoverable. But a ruling on the same day of the Heppner decision by a federal district court judge in Michigan appeared to reach a contrary result. But did it?
The case, Warner v. Gilbarco, involved a pro se plaintiff who apparently used some “AI tools” in some unspecified fashion. Defendants requested the documents from that work and Warner claimed work product privilege and moved to compel. The court denied the motion.
While on its face, the ruling appears inconsistent, on closer review it isn’t necessarily. In fact, the court began its analysis by plainly stating the material in question was not relevant, or even if so proportional. It should have ended there.
Transform Legal Reasoning Into Business-Ready Results With General AI
Protégé™ General AI is fundamentally changing how legal professionals use AI in their everyday practice.
The Warner Ruling
In Warner, the defendants argued that by using the unspecified AI tools, the plaintiff waived the attorney-client and work product privileges. It’s not entirely clear how the plaintiff could assert an attorney-client privilege, nor is it indicated whether the pro se plaintiff is an attorney.
The court instead focused on the work product privilege, noting first that a pro se plaintiff can assert a work product privilege under case law in that circuit. The court then stated that a waiver has to be a waiver to an adversary or result in the likelihood that the material in question will get in an adversary’s hands. While disclosure to any third party would waive the attorney-client privilege, said the court, that’s not so for waiver of the work product privilege.
Warner Is Not Necessarily Inconsistent
In fact, the underlying principles and the holding is not necessarily inconsistent with that in Heppner. First, the fact that Warner was pro se is important. Courts often give pro se parties a little leeway for obvious reasons. Warner had no lawyer to go to to formulate strategy and discuss her case. So Warner used tools that were available. And the work product privilege clearly applied to her. In Heppner, on the other hand, the defendant had a lawyer and chose to ignore him, conducting his own search and discussion. That difference is critical.
Second, the Warner court ignored one factor that the Heppner court relied on: the fact that the GenAI tool in that case kept the material that was generated and used it for training purposes. The Warner court merely concluded without analysis that the Warner material was not likely to get into Gilbarco’s hands. That’s an important factor that the Warner court didn’t address.
What Was Really Going On?
It also can’t be ignored that, to be blunt, the Warner court was simply pissed, which often happens in discovery disputes.
Why do I think that? Here is what the court said:
The motion seeks intrusive post-discovery production based on speculation about what might exist in Plaintiff’s internal drafting process, untethered from Rule 26 relevance, disregarding the heightened protection afforded to opinion work product, and improperly attempting to manufacture a waiver where none exists. At its core, Defendants’ request is a fishing expedition…Additionally, the Court agrees with Plaintiff that the pursuit of this information is ‘a distraction from the merits of the case.’
Pretty strong. The court clearly thought the material in question was blatantly non relevant.
And this at the end of the opinion: “In the end, both sides of this dispute seek to obtain each other’s thought processes, while shielding their opponent from discovery of their own. The Court will uphold the protections afforded the thought processes and litigation strategies of both sides and will order production of neither.”
In other words, a pox on both your houses.
To get there, however, the court took some unnecessary detours that raise some questions.
A Little Wonky
Unfortunately, to get where it wanted to go, the court got unnecessarily a little wonky which detracts and confuses the holding. First it said that GenAI tools are not people and therefore disclosure to a GenAI tool could not constitute a waiver, as a waiver has to be to a person. Not sure I buy that since many legal entities are not technically people but are treated as such in a variety of circumstances. And giving GenAI tools non people status could have serious implications down the road when it comes to liability among other things.
The court then held that the materials were her internal analysis and mental impressions rather than an existing documents or evidence. Not sure about that one either: if materials aren’t discoverable because what they contain somehow makes them neither documents nor evidence, I wasted a lot of time over the years producing stuff.
Neither point was necessary to determine a lack of waiver.
Reading the Cases Together
Be that as it may, here is what we can glean from reading the two cases together. Use of GenAI and whether it constitutes a waiver will depend on the facts. Under both cases, it will require a disclosure and that disclosure must be of mental impressions. It also must be in anticipation of litigation not just idle advice here and there.
But above and beyond everything else, the material in question must be relevant to the issues in the litigation, a fact I mentioned in my earlier article, calling the relevancy showing a “tall order” in most cases. It’s clear from reading the Warner opinion that the court felt the material was so non-relevant that the time spent on trying to discover it was nothing more than a time-consuming distraction. The material in Heppner, on the other hand, was clearly considered relevant by the court.
It’s also clear that waiver is viewed differently for pro se parties versus represented parties, as it should be. If the Warner material was relevant and Warner had an attorney, but still used GenAI tools without her attorney’s knowledge, the result might have been different.
So there are some factual differences that are important. But bottom line, both cases suggest that use of GenAI tools in litigation should be handled with care. If you’re a lawyer and represent a client, it’s still a good idea to advise them that they shouldn’t use GenAI tools without consulting you.
Waiver of privilege is a mine field and can easily happen. Nothing in either Heppner or Warner suggests anything to the contrary. And treating them as inconsistent and thereby convincing yourself that privileges protect what you sent and received to a GenAI tool from discovery would be a mistake.
Stephen Embry is a lawyer, speaker, blogger, and writer. He publishes TechLaw Crossroads, a blog devoted to the examination of the tension between technology, the law, and the practice of law.