Introducing AI discovery: speakers, disclaimer, and webinar agenda
Free. Watch without an account. Sign in to save progress and take the exam.
Up next
What generative AI actually is and how it produces mutable outputsTranscriptThe player shows YouTube’s automatic captions, which are machine captions we do not review. The full transcript is below.
00:00:02Discovery of AI, Prompts, Training Data, and the New ESI.
00:00:08Today's speakers are Daniel Garrie, founder of Law & Forensics Neutral Jams, adjunct faculty at Harvard.
00:00:16Scott Doyen, shareholder at Clausen Miller PC, Karen Silverman, CEO and founder of the Contellas Group.
00:00:24Mari Grace, counsel at Debois & Plimpton LLP, Patricia Antezana, counsel at Reed Smith.
00:00:33The program materials can be found under the Supplements tab. Thank you for joining us.
00:00:37I want to now turn things over to today's moderator, Daniel Garrie, for further introductions and to start the program.
00:00:44All right. So hopefully everybody can see the slides as I can see them. So that counts for something.
00:00:53So without further ado, this panel, a couple of our panelists are literally in hearings and are going to join us on the tail side of the presentation.
00:01:03This fascinating area, discovery of AI, you know, I never thought anything I ever did in my graduate work would ever find its way into the law.
00:01:11But prompts, training, data, and all this information is something lawyers really need to become familiar with at some level,
00:01:20at least both client facing and probably internally as well. So we have a fantastic panel.
00:01:26I'm just going to let my panelists introduce themselves. They're all world class within their own rights.
00:01:31I'm just going to read the disclaimer, then let them introduce themselves.
00:01:34This is not legal advice, nor should it be considered legal advice.
00:01:37This presentation of the comments contained therein represent only the personal views of the participants as spoken and do not reflect those of their employers or clients.
00:01:46This presentation is offered for educational, informational uses only.
00:01:49So I will let Karen and then Mari introduce themselves, and then we will kick it off.
00:01:56My relevant experience just quickly, I wrote the AI rules for resolving disputes on AI platforms for jams, and I've been working in technology.
00:02:09I mean, I don't consider myself an expert, even though my master's is in it.
00:02:13It was 20 years ago, so I'm not foolish enough to say that I am as smart as any of the people I work with.
00:02:20But I do work in AI on a regular basis.
00:02:24So with that said, Karen, I'll let you introduce yourself, then Mari, Karen, the floor is yours.
00:02:29Thank you. I will say that I always learn something from you, so I'm going to call you an expert.
00:02:35Karen Silverman, founder and CEO of the Kintellus Group, we really focused on technology governance and tell us legal is a piece of that.
00:02:45So we are in and around these issues all the time, really from the perspective, though, of how to manage getting the kinds of evidence I think we're going to be talking about here today.
00:03:01I sit on the ABA AI task force and have done a variety of other sort of policy related jobs and have had those roles.
00:03:12But at the moment, most interested in where the rubber meets the road.
00:03:16So very happy and looking forward to this conversation.
00:03:22Former partner at Latham. So sorry about that.
00:03:25I mean, general rock star. There's a whole bunch of other things she's done.
00:03:29I forgot. No, it seems like a whole chapter ago. It was a whole chapter ago.
00:03:32Anyway, I was a partner in the competition practice at Latham for many, many years.
00:03:37Thanks, Karen. I'm Mari Grace. I'm a counsel at Debevoise in the AI Group, which is part of our Data Strategy and Security Group.
00:03:46I am new to Debevoise and focused exclusively on privilege, discovery, governance, all manner of litigation risks surrounding AI.
00:03:56And before that, I was a civil litigation partner at Davis Polk, where I was really since law school and focused very much on the intricacies of discovery in all sorts of federal and state cases.
00:04:10All right, so we're going to cover overview of AI, AI data as ESI, requesting AI artifacts, because I think lawyers really struggle with that.
Automatically generated; it may contain errors.
Study materials
Notes, flashcards and practice questions for this lesson, reviewed before publication.
Study notes
Why prompts and training data now matter to lawyers
The course opens by framing a shift: concepts once confined to technical graduate work, such as prompts and training data, have found their way into the law. Daniel B. Garrie says that prompts, training data and the information around them are something lawyers really need to become familiar with, at least at some level. He names two settings where that familiarity matters: client-facing work, and probably a firm's own internal use of these tools as well.
Three subjects the AI discovery panel set out to cover
Daniel B. Garrie outlines the discussion in three parts:
- An overview of AI. The starting point is what these tools are, so that the later discussion of their data rests on a shared understanding of the technology.
- AI data as electronically stored information (ESI). The premise behind the phrase "the new ESI" is that prompts, training data and related records are a category of electronically stored information that discovery has to reckon with.
- Requesting AI artifacts. He singles this topic out because, in his words, lawyers really struggle with it. Knowing that AI data may be discoverable is a different skill from knowing how to ask for it.
The agenda moves from general AI concepts, to the status of AI data in discovery, to the practical problem of drafting requests for it.
Managing AI evidence where the rubber meets the road
Karen Silverman frames the issues the panel will discuss from the perspective of how to manage getting the kinds of evidence at stake in AI disputes. She says her main interest at the moment is "where the rubber meets the road": the practical handling of AI evidence, rather than policy questions alone. Technology governance, in this framing, includes the practical question of how evidence is obtained from AI systems.
Privilege, discovery, governance and AI litigation risk
The introductions also name the issue areas that run through the rest of the course: privilege, discovery, governance, and what is described as "all manner of litigation risks surrounding AI." These are the areas in which AI data creates new questions for lawyers. Later lessons take them up in turn: what generative AI is and why its outputs vary, the vocabulary of prompts and model layers, which discovery rules govern AI data, what AI evidence exists and why it disappears, who possesses it, how to request it, and how privilege and work product apply to AI use.
How this framing prepares the rest of the course
The framing gives a learner three questions to carry through the course: how the technology works, whether its data is discoverable as ESI, and how to ask for that data in a way the other side and the tribunal will understand. Each later lesson answers part of one of those questions.
Flashcards
8 cards. Test your recall, then flip to check.
Select the card, or press Space, to flip it. Left and right arrow keys move between cards.
1 of 8
Practice quiz
Practice — not graded, not the final exam