In Episode 164, we celebrate our 4 year anniversary of Case of the Week! Stephanie Wilkins of LegalTechnology Hub and Aaron Patton, aka Gates Dogfish of Trustpoint.One join Kelly Twigger to look back on the last four years in ediscovery and legaltech and what we can expect on the horizon.
Introduction
Kelly Twigger: Hi, and welcome to this week’s episode of our Case of the Week series brought to you by eDiscovery Assistant in partnership with ACEDS. We are incredibly excited to have you join us for this final episode of 2024 and to celebrate our four-year anniversary of the Case of the Week. Amazing. I can’t believe it’s been four years. My name is Kelly Twigger. I am the CEO and founder at eDiscovery Assistant and the Principal at ESI Attorneys. I’m so happy to be welcoming two of my colleagues from the industry today, very special guests, Stephanie Wilkins and Aaron Patton. Thanks guys for joining us.
Most of you know that when I invite guests to our Case of the Week, I always try to bring different perspectives to add to my thoughts, and this week is no exception. Let me introduce our two guests. Most of you likely know Stephanie Wilkins. She is the former Editor and Chief of LegalTech News at ALM and is now the Content Director for the Legaltech Hub. Stephanie was also a practicing attorney in big law before moving into journalism, so she knows and understands the issues that lawyers are facing every day.
Her outstanding work in journalism for the legal tech space was recognized last year when she was included in the FastCase 50, and again, this year as the recipient of the Inaugural Journalism Award at the Legal Technology Awards. She and her team did incredible work at ALM and really elevated the coverage of our legal tech news exponentially. We certainly miss you in that role. I think my favorite thing about Stephanie, and I took this from her LinkedIn, is that you’re a boxer in training. I love that. Thanks for being yourself.
Stephanie Wilkins: Thanks for having me.
Kelly: Absolutely. Next up, celebrity persona in the ediscovery space, Aaron Patton, who we all know as Gates Dogfish, the publisher of incredibly hilarious and spot on memes on social media related to ediscovery. As Gates Dogfish, Aaron has published just under 600 memes on social media since 2022. He’s an attorney and the managing director at TrustPoint, a well-known service provider in the industry. He’s got a ton of technical knowledge to share with us today, including some fabulous memes. So thanks for joining us, Aaron.
Aaron Patton: It’s my pleasure. Thank you.
Kelly: One quick announcement before we dive into our topics for discussion. Coming up on February 12th and 13th, 2025, is the University of Florida eDiscovery Conference. This is an incredible opportunity for free CLE from some of the best in the business. I’ll be spearheading our case law panel to kick off the event, and Stephanie will be joining us this year to give a lightning talk. Stephanie, what’s your talk on this year?
Stephanie: Preserving prompts. Will prompts become part of ediscovery? Should you be preserving them? Who’s preserving them? Will they be discoverable? And then will they be potentially privileged?
Kelly: That’s a huge issue that I think is going to come up more, and we’ll cover a little bit in our discussion on Gen AI today. Here is the link to register. We’re already approaching a 1000 registrants. Last year, we had more than 3,500 folks sign up to view us virtually as well as a couple of hundred in person. So please join us.
All right, let’s dive into our discussion today. Today’s episode marks our 164th episode with 39 of those in 2024. It’s incredible that the pace of case law in ediscovery continues to evolve so rapidly. We’re approaching 40,000 decisions in eDiscovery Assistant with more than 4,700 decisions this year so far. On our previous anniversary episodes on which we welcomed David Horrigan and Doug Austin, we’ve done more of a summary of decisions that we’ve seen on key topics. While we’ll do a little bit of that, I wanted to back out a little bit and look at the bigger picture as we end 2024 into 2025 and really see what’s happening in the industry on some of these key topics. We spend so much time discussing them, but what’s really happening? And so Aaron and Stephanie are going to give us those perspectives as well today, and I think that’s just incredibly important.
Before we dive into some of those meaty topics, though, you guys both had one issue that you said still seems to come up all the time, and that is the ever controversial dash or no dash in ediscovery. Stephanie, what are your thoughts?
Stephanie: I personally don’t like the dash. I did a poll on LinkedIn a month or two ago. I gave three options. I gave the big D and little D with no dash option. But if you combine those two together, 77 % went for no dash and only 20 % went for the dash. I think there’s something interesting to talk about in the three percent after we break down Aaron’s opinion.
Kelly: I’m firmly in this dash camp and have been since 2006, probably.
Aaron: Same. The memes that seem to get so much attention are these nitty-gritty niche memes, like the one that we’re talking about right now. I just had one up from Futurama where he’s squinting at people who use a hyphen in ediscovery. And Martha McPherson’s most recent one, her inspiration for it was a Christmas story. I’ll share that one real quick. And It’s poor Ralphie getting soap in his mouth for using ediscovery the wrong way.
Kelly: eDiscovery the wrong way with a dash, I might add.
Aaron: Exactly. This is how he was taught that it’s eDiscovery, lowercase E, uppercase D.
Kelly: Aaron, you mentioned there was something that came out of the Georgetown conference that was compelling on this issue.
Aaron: Yes, it was interesting. When talking about ediscovery, electronic discovery, e hyphen Discovery, one of the things that jogged in my memory from Georgetown was that someone said that we’re reaching an inflection point where more people will have been working in ediscovery software, still growing up in ediscovery software, than the people who are around from the paper days who bridged the gap into electronic discovery. And it was a really an interesting thought because we were joking around about slip sheets and I think they’re called placeholders now when you’re doing it in a production. But there was a time when we were dropping pink slips of paper between documents and then green slips of paper for families, manually affixing Bates’ stickers. There’s going to come a time when people won’t know what the DII file is. They could look it up, but it’s coming when people won’t know.
Kelly: People ask me this all the time, Why ediscovery versus just discovery isn’t all discovery electronic? Well, 99% of it is electronic. But the difference for me is that ediscovery is a whole new set of rules and issues that we didn’t have with paper discovery. And is it all part of discovery? It absolutely is. But the issues that we have are unique to ESI. Therefore, in my view, they’re ediscovery issues. That’s my thought on the dash versus no dash and complexity of ediscovery versus just discovery.
Stephanie: I will say the 3% that was neither or the write in, they were all universally just dropped the E altogether. It’s just discovery.
Kelly: And you know what? I feel like if people actually knew what they were doing when it comes to ediscovery — all across the board. Obviously, there are lots of people who do know what they’re doing. There are a lot of people who still don’t, and we see that all the time in the case law. I think when we get to the point where people understand what the issues are in ediscovery, then maybe it does just become discovery, and that language is just language.
Hyperlinked Files
Kelly: Let’s turn to another big topic from this year, and that is hyperlinked files. I know I feel like I’ve talked this issue to death. I’ve done two webinars, if not three, multiple presentations on the hyperlinked files issue. I think we were going to kick this off with the general discussion. Steph, you raised the question of, what do we even call these? You want to tackle that?
Stephanie: I’ve seen a lot of discussion out there, modern attachments versus hyperlinks. I think calling them hyperlinked files or hyperlinked attachments covers it, because if you just put hyperlinks, people don’t necessarily know you’re talking still about files. Modern attachments, to me just is fluff and meaningless. I also think that this is an issue that even the fact that we’re debating what to call them, I’ve seen so deeply in the ediscovery community, and I think outside of it is still just a mystery. I think if I picked somebody who was not in ediscovery at all and said, what do you think a modern attachment is, or what are we doing with hyperlinks? They would be so not anywhere along the spectrum of what we’re talking about.
Kelly: Totally agree. Aaron, how do you guys address it at Trustpoint?
Aaron: We have some workflows to address it, but before I get to that, I think one thing that is really interesting is that all of the context in which I’ve seen the discussions has been about hyperlinked files in communications. But those hyperlinked files are everywhere. They’re in Evernote, they’re in DocuSign, they’re in Salesforce. So my prediction for 2025 is we’re going to start to see case law about these hyperlinked files that’s outside of messaging and outside of communications. And in that scenario, are they attachments? We’ll see.
Kelly: I think you raise a really important point. At eDiscovery Assistant, we call them hyperlinked files. The reason for that is they are files that appear as a hyperlink, as opposed to being physically attached to the document. There’s a really good article that Doug Austin just published today on his blog from Cloudficient. It gives a really good discussion about the variety of things that people refer to these hyperlinked files as and what really the difference is. There’s a huge difference, and I’ll just bridge that into the case law discussion, which is the first time that we really talked about hyperlinked files as discoverable documents was in the Nichols v. Noom case in 2021, and that one was from Judge Katharine Parker. It was a unique set of circumstances because the parties had agreed early on — they’d recognized that there were these hyperlinks. The parties agreed that a certain number of them would be provided. Then Noom went ahead and produced its documents out of Google apps with hyperlinks.
It turned out that there were almost 10 times as many hyperlinks in the collection that was produced as the plaintiffs anticipated. They went to the court and asked Noom to go back and reproduce them. The court made this pronouncement where she said hyperlinks are not attachments. That was problematic in the ediscovery space, as you might expect, because when you have a hyperlinked file that is attached to a responsive message or is otherwise responsive under the Federal Rules of Civil Procedure, it likely should be produced. Then we evolved a bit in the thinking to really starting to distinguish from hyperlinks to hyperlinked files. If you watch our eDiscovery Day webinar you’ll see a whole discussion of this case law, so I’m not going to spend as much time on it today. But from Nichols, we moved to the In re Google consumer privacy case. In that case, the court really looked at the volume of hyperlinks and really took a sampling approach for it and ordered a specific number of documents to be produced at hyperlinks based on some information from plaintiffs that suggested they were key and relevant.
We then moved to the In re StubHub Refund Litigation in April of 2023. That litigation, StubHub agreed to provide links or the documents at hyperlinks and ultimately determined that they could not. They put it in the ESI protocol that they would do it. The court ordered, “hey, you guys put this in your protocol, you have to live up to what you put in your protocol. An agreement is an agreement.” Some really great language. I believe most all of these cases we covered on Case of the Week, so they’re accessible via the blog as well. In re StubHub then came back a year later to the court and said, hey, we’ve continued to try and we’ve spent a whole bunch of money doing it, and we cannot pull all of these documents. We can’t get the contemporaneous versions, the technology just isn’t there to allow us to do it. And the court said, well, your ESI protocol says you can be released from obligations here for good cause shown. And if the technology isn’t available, that’s good cause. So the court really let them out of that.
The In re Uber Techs., Inc. Passenger Sexual Assault litigation is my favorite decision on the hyperlinks issue. That one’s from Judge Cisneros, where she really required the parties to sit down and have discussions about what was technologically possible to collect documents from hyperlinks at Uber. What’s important in all of these cases is that each party has a different technology where hyperlinked files are used and the technologies that existed at that time to collect documents from those hyperlinks are all different. So you really have to pay attention to what are the technologies where the hyperlinked files are created and stored and what technologies exist to allow us to go and get them, and what do those technologies really allow us to do? So can we collect a contemporaneous version? Is there any ability to create metadata to associate the hyperlinked file with the original parent document? And what do we even call that parent? We still don’t really know. There’s a subsequent In re Uber Techs., ESI protocol that the parties entered into with language that Judge Cisneros has ordered in that previous decision, which is excellent language and one that I routinely recommend folks look at as a baseline for evaluating how to address these in your own ESI protocol. I highly recommend that one.
There are other decisions. There’s a couple of decisions in which Meta was involved, in which at one point Meta said, and I think it was the UAB Planner case, Meta said, We can’t provide all these documents, but we’ve reviewed them and they’re not relevant, which was interesting because if you managed to review them, how can you not provide them? But Meta also had an earlier decision in the In re Meta Pixel case where they managed to convince the court that even though they had Microsoft Purview, which had some functionality at the time to allow them to collect hyperlinked documents, they claimed that it would interrupt their ediscovery workflow, and so they somehow managed to to convince the court that they should not be required to provide hyperlinked files.
Aaron: I love these because it’s almost like Obi-wan Kenobi, “These aren’t the docs you’re looking for.”
Kelly: Oh, my God, I’m such a huge Star Wars fan. I’m going to use that going forward!
Aaron: But what’s also so fascinating about this moment is we’ve got reality in conflict with the desires of the parties or stipulations of the parties. It’s so interesting. And it reminds me of those moments where it’s like, you were supposed to process these documents and give us custodial information, get it from the documents. It’s like custodians are assigned. It’s not extracted. And so let’s make sure we know what we’re talking about here because that’s not possible.
Kelly: You raise a great point. And Aaron, you did something recently on hyperlinked files on LinkedIn that you posted your deep fake watermark that I want you to talk about. But just to reference again, that eDiscovery Day webinar, we talked about some of this, so I encourage folks to go back and listen to that. What was the thing that you posted on LinkedIn about this issue?
Aaron: It was a thought experiment. So taking some of these cases and just thinking, what would it mean if a corporation’s electronic communication policy became, you can only use hyperlinked files. You’re not allowed to attach documents so that there is no parent-child relationship. And you have to get permission from the legal department to do it. You have to get permission from IT. It was a joke, right? And it was subject to the jester’s privilege. It has this watermark that says Super Fake.
Kelly: We never know. We’ve moved away from physical attachments. Now we’ve moved to hyperlinked files. I mean, you could ultimately preclude physical attachments. You could do this.
Aaron: Now, why would you do it? Because I was thinking, this is absurd. But then, putting on my lawyer’s hat, I was like, wait, but if I send somebody an attachment in an email, if they forward that to somebody else, that person can open it if it’s not password-protected. But if I send a hyperlinked file, the access controls on that will keep them from opening it. That’s actually interesting. And then there’s also storage size, right? You’re not going to bloat your mailbox. It seems like there are legitimate reasons for doing this, And so long as you’re not intentionally frustrating discovery with this policy, I could see somebody trying it.
Kelly: And Stephanie, just to kick that over to you, like you mentioned, there’s a lot of implications on the hyperlinked files issue for the knowledge management world.
Stephanie: Well I don’t think your scenario is that far-fetched, actually, because if you think about, even now, if you operate Microsoft 365 and I send somebody, I almost never attach a word document. It’s just a link to go find it in OneDrive or SharePoint or whatever. It’s the norm, but it’s still exactly in line with how I used to send a file. I’m not doing anything in my brain any different. I’m like, here’s a word file. So I don’t think that’s that far-fetched. I mean, yeah, putting in the policy that we’ll only do that is a little bridge further.
Aaron: So you only use links?
Stephanie: Well, no. I mean, not specifically intentionally, but most often I feel like I end up sharing links these days.
Aaron: Microsoft defaults to a link, and I always undo it and attach it. So I do the opposite, which is interesting.
Stephanie: It’s because it defaults. I’m like, well, that’s fine. And I mean, I’m not doing anything, knock on wood, that has any, hopefully any discovery implications on the road. It’s just like with my content team, like here’s a draft, review this article. But it’s just as the technology progresses, these hyperlinks are just part of what we’re doing. So to ignore them is just contrary to reality of technology right now.
Kelly: And I think it’s really interesting because we don’t think of hyperlinks on the same level as a generative AI. I was listening to Brené Brown’s podcast yesterday, and one of the things they were talking about was generative AI, and she was commenting that this is what happens when technology moves faster than policy. And that is, it feels like that’s where we live 100 % of the time in ediscovery.
Aaron: That’s reality.
Kelly: And Hyperlinks are a very solid example of that situation. In my view, you look at the Federal Rules of Civil Procedure and a party’s obligations to provide documents under Rule 26, those documents should be provided. If you’ve created technology that doesn’t allow you to go and get them, and you’re a serial litigant — these are some of the points that Judge Cisneros has addressed in the In re Uber Techs. case. You have to be prepared to provide them. You have some responsibility for the technology that you choose and the capabilities that you lock yourself into. If that requires you to do more manual work, which it can be a lot of what’s required to get documents and hyperlinked files, then that’s what you’re on the hook for. It was interesting because she was very willing to evaluate the party’s obligations in the scope of the technology, but also in the scope of what their requirements are under the rules. She’s the first Judge I’ve really seen do that. I think we’re going to start to see technological developments here. I’m not a huge prediction person. I know you guys both have your predictions that I want to talk about, but I really think we’re going to see…I thought by this time, this year, we would have seen the improvements to allow us to do things like that.
Microsoft did just roll out some changes where that if you implement a retention policy in Teams, anytime you put a legal hold in place after that retention policy is in place, the documents and the hyperlinked files will be preserved for messages that you have put a hold on to that date of that retention policy. But if your data goes back earlier than that retention policy, you’re not going to have them. It’s going to take a while for the technology to catch up with what are our traditional preservation obligations. Cases are won and lost in civil litigation on the documents. How are we not providing the documents at hyperlinked files? It’s got to be fixed.
Stephanie: I also question, you mentioned that setting in Teams or whatever, that these are optional settings you can put on right now and people can just willing, I mean, whatever level of willfulness is there, choose not to retain these documents when they know maybe they should because they’re not required to. At what point is that balance going to tip where it’s just you’re doing something wrong if you’re not putting in place these options that are right at your fingertips to put them in place?
Kelly: I would agree. I think it’s going to be a question of what is your obligation to preserve and does it mean if you have the capacity in-house to trigger that functionality and you don’t do it? Then again, what some of the case law seems to suggest now, what the In re Uber Tech’s case with Judge Cisneros is, look, if you know that you have technology and you know you’re a serial litigant who’s going to have to produce information, you need to take the steps to be able to be prepared to produce it or to do whatever manual work is required. Aaron, you had a comment.
Aaron: Well, one bit of the nitty-gritty was that I think we all agree that responsive documents that are in scope, which many of these hyperlinked files would be, should be produced if they’re responsive, nonprivileged. But I think one of the hang-ups here is the creation of a family relationship between the two. Because the software, it can extract an attachment from the MSG wrapper or the EML wrapper in an actual email message, and you see that it’s in there, which, by the way, is one of my favorite little things to point out, is that if you’re hosting the native email and all the attachments, the attachments are still in the native email, so you’re hosting them twice in a lot of scenarios. But what’s interesting is those relationships are embedded within that file. And so if you have to actually go and create a relationship between something that’s over there. At Georgetown, Paul McVoy, who I think was part of one of these cases, he was saying it’s a simple script. That’s what it is.
Kelly: I don’t understand why it’s that complex to create a metadata field. We do it all the time. We sure do. Create custom metadata fields for data because we know that there’s one metadata field in a particular case that we really need to see, and we just write a little script to create that metadata. I mean, we do it with redaction. People didn’t used to do this. We ask for in our ESI protocol, a redaction field that’s populated, yes or no. We can search, we can filter by yeses, and then people can review those redactions for privilege out of the gate. One of my big soapboxes is that you’ve got to review your productions when they come in. The longer you sit on them, the less ability you’re going to have to be able to challenge anything when the courts are so pushing parties to move everything forward quickly now.
Aaron: So Kelly, I’m sharing a meme right now of Winnie the Pooh, and he’s…
Kelly: The overlay is my friend.
Aaron: Exactly. I actually have one that I made with AI that has a beer that has an overlay label on it. “eDiscovery giving you a hard time, crack open an overlay.”
Kelly: Yeah, well, that’s about right. If you don’t know how to use overlays, you have to find tech people that know how to help you. The best thing that I found is when you’re an attorney and you say, Here’s what I need to be able to see. You don’t need to tell them technically exactly what needs to be done. If I came to Aaron and said, I need a metadata field that’s going to show me this. I know we have that information, and you go, Okay, great. No problem. I just write this quick script, I overlay it here, and suddenly I have a field that’s populated. So lawyers, tell your technical people what you need and let them help you identify the technical solution to get what you need.
Aaron: Yeah, that’s 100 %. In working with my tech team, I tell them the result I want. I don’t tell them how to do it because it would take me forever, and the way I’m going to do it is so clunky.
Kelly: Exactly. And they don’t need to know and they don’t want to know. That’s the whole purpose of the relationship between legal tech support and litigators is they need your help to be able to see and get the information organized the way that they need it.
We also talked a little bit coming into today about the impact on ESI protocols for hyperlinked files. Stephanie, I thought you had some comments on this about folks that you’ve talked to as well as what we expect to see in the future.
Stephanie: I’m trying to remember what I said now. That is embarrassing.
Kelly: Oh, sorry.
Aaron: It was so good.
Stephanie: I know. It was like a bit of genius, and I don’t remember it.
Kelly: I think it was along the lines of this is yet another issue that you’ve really got to be on top of going into your ESI protocols, and you can’t wait until you’re trying to negotiate the ESI protocol to learn it. It’s yet one more reason why you have to be understanding your client’s data sources early on in a matter or before a matter. We’re back to needing data maps again, needing refreshed data maps.
Aaron: It was templates, Stephanie. I think we’re talking about templates.
Stephanie: I’m glad everybody remembers what I said. But I really was struck by that quote you said a few minutes ago about the technology always moves faster than policy. I feel like that should be the motto of the ediscovery industry, because I’ve been saying forever that ediscovery professionals are the ones that are always on top of how technologies work before most people because they have to be. So again, and this is what we’ll get into, well, what I’ll get into at UF, about conserving prompts and things, too. You just need to be thinking about these things. And it’s harder to go back and add stuff in. Just think about it now. If you know this is coming down the road, put it in your policies now, put it in your protocols now. It’s like trying to shoehorn things in later. Like you were saying, review documents early. Don’t wait. It’s harder and harder to find things.
Kelly: Know what your capabilities are. Look at the tools that are out there in the market. If you use Microsoft Teams, look at Purview. One great change that was made, I’m not sure whether it was this year or last year, was that Microsoft now allows you to have only a few licenses on E5. It used to be You had to pay for the whole organization on E5 if you wanted E5, and that was so cost prohibitive, no one would do it. Microsoft wants people to use their ediscovery tools that they’re building, and they’re continuing to iterate on it. There was a great discussion, I think it was Tom O’Connor and Rachi Messing and Stephanie Biers from Microsoft, where she talked about all the work that they’ve been doing. Of course, Rachi used to be at Microsoft, and so he and Stephanie had worked together. It’s a great discussion. It’s really worth listening to. There’s a second one with Arman Gungor, he built Forensic Evidence Collector at Metaspike. It’s called FEC. That’s the tool that’s actually mentioned in the Noom case back from 2021, and other folks have mentioned it.
Kelly: That tool allows you to collect hyperlinked files from Google, but not from Google vault, which is the problem in the Uber case. So again, you got to know your technology, you got to know what the capabilities are. So go out and learn these things. If you’re dealing with clients and you don’t get the information from the client up front, know what’s available and out there and ask the questions. What are the technologies you’re using? How are you archiving things? Is there a combination of Google Vault with Microsoft? What are we doing? So we can really understand. You can at least have a baseline knowledge, but you have to spend some time in advance.
Stephanie: And even give them a list of popular ones, because they might not even know all the technologies they’re using. But if you give them a list, they might be like, oh, yeah, we do use that.
Kelly: Right. What capabilities are there? I mean, ask, does someone have access to E5 on your Microsoft teams? The answer may be no, but counsel are going to have to go to IT to learn those answers. I get a lot of questions from folks always saying to me, Oh, how are you — you know so much as a lawyer, how are you so technical? Because I just ask questions. I’m just not afraid to ask questions. I’ve had my share of IT people go, “what are you talking about?” I don’t know. Why don’t you tell me what I’m talking about? Here’s what I’m trying to solve. Here’s what I need to know. You talk to me. Then they talk to me and I type as fast as I can. Then I ask follow-up questions. You have to learn. You have to be dumb until you get smart. Every new technology requires you to start over again.
Aaron: Kelly, I’m going to share a meme for you.
Kelly: Great. “Didn’t learn ediscovery in law school, huh? They said we’d have people for that.” It’s interesting to me because I just taught at CU last year for the first time, and I’ll be picking that back up again. But everyone says that this generation, because they grew up with phones and technology, that they know ediscovery. No, they don’t.
Aaron: So not true.
Kelly: No, they don’t. They don’t have any clue. They know how to create, store, send, and receive electronic information in a variety of ways. Text messages, Signal, WhatsApp, Yammer, TikTok, whatever you got. But they don’t think about it as evidence or how to go and get it, which is exactly what ediscovery is. It was fascinating. I stood up the first day and said, Okay, guys, tell me some sources of ESI. They said, email, nobody said mobile phones. I held up my phone and I was like, Hello, what about the phone?
Aaron: Everything on this.
Kelly: Well, that’s important. What do you mean? I think any misnomer we have that the younger generations know and understand ESI because they’ve been creating electronic information is a false narrative.
Stephanie: They’re just really great at creating data. They’re probably more prolific in creating data than anyone ever. But what happens to that data or even thinking of it as discoverable data is a completely different question.
Kelly: The kinds of data they create, too, is what is problematic for us. We’re used to text, we’re used to word documents, we’re to things like that. All this audio, all this video, it’s expensive. You have to OCR it, you have to have it translated, you have to have all these things. It’s much more complex now to handle a collection of information than it was five years ago, even.
Aaron: Although, you know what’s really cool? Just something I learned recently with Dropbox, I have all these pictures from the last, I don’t know, 15 years uploaded there. I was looking for a specific picture, and I didn’t know what it was called because I never named it. Because of AI, because of computer vision, it can see what’s in the picture. And I searched for Tuxedo, and it found my prom picture so that I could show it to my son, and we look exactly the same. He’s 13, and I was 16, but we look exactly the same. But I didn’t have a name for it. I was just searching for what was in it, and it came up.
Kelly: I do the same thing on my iPhone, right? My husband and I were picking out pictures for the Christmas card, and I’m like, Oh, I had this great picture, whatever. And so I searched just using the name or whatever on my phone, and just like that.
Aaron: That’s because we’re search term people, right? We grew up with search terms.
Stephanie: I did it the other day. I was trying to find a picture of something I once saw in a museum sometime in the last two years. And then I just typed “doll” in my photos, and it brought up all these options. It was one of the first five. I was like, oh, yeah.
Kelly: Because it creates all this metadata and then it goes and looks at things and adds more metadata to your photos.
Generative AI
Kelly: I think that’s a perfect segue to our next topic, which is the topic of all topics, generative AI. And just to start, because Steph, I want to turn to you for thoughts because you had a lot. Just to be clear, we’ve been using AI for years. If you get Netflix recommendations, if you get podcast recommendations, if you get anything that’s recommending stuff to you based on choices you’ve already made, that’s AI. Generative AI is when you ask AI to create something for you, whether it’s text, a video, audio files, an image, whatever, and it’s actually generating that content for you. And the generative AI is what’s really big. And this is yet another example of technology outpacing policy.
Stephanie: Faster than ever.
Kelly: I know. I’m going to turn that over to you Steph.
Stephanie: Yes, AI is far from new, and it’s not even anywhere near new in legal. This Gen AI boom has just put it on a lot of people’s radar that we’re thinking about it. But the technology itself is moving so much faster than other technologies before it even. If we couldn’t get policy to keep up with technology before this, now it’s this light speed ahead. I mean, it’s breaking all the rules of how fast technology progresses. So we are really in a world where people are just using this because it’s also so much more accessible. The big thing ChatGPT did, and it’s not a legal tool, I always have to say that, but the big thing it did is put this technology right in people’s hands in a way that hasn’t happened before. It was like every single person with an Internet connection can suddenly play around with Gen AI and feel the magic and just get swept up in it. That was actually brilliant on OpenAI’s part. Obviously, I don’t agree with everything they ever do, but that was getting it in people’s hands and making it so easy to use and get people to buy in. It was so fast.
Kelly: Just for people, Steph, who maybe aren’t as familiar with or are very overwhelmed with the discussion of generative AI, and they’re like, Stop talking about it. I don’t want to hear about it anymore. How do you use prompts in ChatGPT? Because that’s your discussion at UF. How do you use prompts in ChatGPT to create content.
Stephanie: So no matter what tool you’re using, very generally speaking, prompts work the same way. There’s just a little box. You type it in like a picture. It’s like the chat on the side of this. You put it in a little box like that, and you tell it what you want it to produce. And the more successful, where prompt, and for a while, prompt engineering was the big term, is that you can put more and more detail in your prompts. You could be like, give me a picture of a flower, and who knows what it will give you, or write me a poem. But if you say something like, I want a picture in a wide screen aspect, in watercolor style, and it should be pink in these, whatever, you’ll get a totally different response.
Kelly: If you say, I want an orchid that’s pink and white, and I want it in this setting, and I love the watercolor view. And it does, and it can be amazing what can be generated. But policy-wise, there’s so many problems with it. These are a lot of lawsuits that OpenAI is facing is they’ve scraped a lot of data off the internet, some of which is copyrighted. I’m not speaking to the merits of the litigation at all because I haven’t paid attention to a lot of it. So please do not take this as any basis of that. But that’s the challenge. Artists who have created visuals are being uploaded and people will use prompts to ask for changes to that art. And there are a lot of questions about copyright violations associated with generative AI, which I think is a huge problem.
Stephanie: I was never an IP lawyer, but modern IP law doesn’t fit with Gen AI. It just doesn’t. So it’s either people are trying to make existing law apply to something that is just not… It’s like a square peg round hole, or this has a potential to completely rewrite copyright law, trademark law.
Kelly: And I would say even that statement of policy moving faster than… Or tech moving faster than policy, moving faster than the law, which is not surprising. I’ve always felt like ediscovery, the law and ediscovery and the changes to the rules and things have moved a lot faster than anything else we’ve seen in the law. But we’re just in such a political gridlock at the moment in Washington that I don’t know how we get the policy we need on AI in the United States. What were you going to say, Aaron?
Aaron: Well, I was going to jump in with a quick meme to explain prompts.
Kelly: Great.
Aaron: And think of Harry Potter with the book in Chamber of Secrets. Right?
Kelly: And he writes, My name is Harry Potter.
Aaron: And ChatGPT writes back, Hello, Harry Potter. My name is ChatGPT. In the book, it was Tom Riddle. But one of the things that’s really interesting about the prompting, and what it can do is it keeps changing also. So I did a prompt engineering course through Coursera with Vanderbilt University, and ChatGPT 3 was what they used. And in this scenario, when you’re watching what the professor is doing, it can actually output a PowerPoint. Make me a PowerPoint, and you have a PowerPoint, you can download and open. It doesn’t do that anymore. Now you ask it for a PowerPoint, and It will give you the text that you can put on your slides unless you know how to prompt it, in which case, and I think I’m going to make a video about this, where you ask it to create the visual basic encoding, and then you drop that into a macro, and you can tell it… here’s your text, I want you to find images, I want a design that has high contrast, that has technology motifs, whatever. You give it the description, it will actually, if you wanted to, it’ll do the research. But if you already have something, you drop it into the macro, you hit run, and it creates a beautiful presentation with moving backgrounds, with the motifs you’re looking for, the color schemes you’re talking about, your text. It’s one of the ways that I’ve been making macros for Excel for analyzing DAT files, which I have another one coming soon. So it’s this really interesting thing where it’ll program in five minutes. It’ll give you the text you need to run a macro to find bates gaps in a DAT file.
Kelly: That’s just another example, I think, of how you leverage your tech folks. I mean, one of the things that I talk about a lot is this gap between the lawyers and the tech folks. The tech folks have so much potential for helping, but they don’t know what’s happening in your case or what the problems are you’re trying to solve or what you’re trying to do. If you would bridge that gap by having the tech folks involved with you from the beginning, you can solve things like knowing what you can put in your ESI protocol about hyperlinked files or metadata or whatever. You can solve this problem right. You could say, Hey, I need a really fast PowerPoint to go to the client tomorrow because that comes up all the time. And how am I going to do this? Well, you call Aaron and he runs this quick macro. So call your folks. Get on with them and let them help you leverage all all these tools. All this discussion that we’re having is not to say that you need to be the person who knows everything, but what you need to know is what the capabilities are and who you can call to make it happen.
And it’s just there’s so much opportunity with AI. Now, that being said, I will tell you that my view on generative AI is there are some amazing applications for it right now. We’ve used ChatGPT to create the summaries of all the 40,000 decisions in eDiscovery Assistant. It’s huge. It’s something our users had been asking for for years, and it was just not possible to do it on a one-off basis. That being said, we used ChatGPT 3.0, as Aaron was talking about. Then it’s subsequently morphed to 3.5 and then 4.0. And wow, the summaries that we get now are 100 times better than the initial ones that we got. We’ve gone back and rerun a lot of them because it’s better, it’s more useful, it’s helpful, and it’s also so inexpensive to do that it’s a huge value add. I think as the reasoning gets better with AI, it’s going to have more application. But there are some incredible things right now. Identifying PII. I mean, we had some tools to do that, but the Gen AI tools that do that are so great. Tackle AI is doing some incredible stuff in that space. I just think there’s so much, so much ability. Aaron, what are your clients asking for or using when it comes to Gen AI?
Aaron: So I’m glad you asked. I’m going to share a meme. And this meme, for those who don’t see it. It’s the Joker and Pennywise talking about all the existing analytics tools gesturing off in a graveyard. So we’re testing a lot of things.
Kelly: I think that’s key. All the tools that nobody’s using.
Aaron: And that’s part of my answer, actually. And so we are testing generative AI tools, but what really drives our use, our implementation of them is client demand. Right now, we don’t have any. And we bring it to people, and they’re like, we’re not ready for that.
Kelly: So is it a cost? Is it trust? Why? Do you know?
Aaron: Well, the cost is not insignificant. I think that people are worried about being that first reported case. That’s not nothing.
Kelly: No, it’s not. You got to have somebody who understands the technology well enough to make a solid argument to the court. You have to have a lawyer who can make that argument to the court.
Aaron: And experts who are going to be able to sign those affidavits and give testimony and go in that battle of the experts. So back at my old company, my colleague was one of those experts on the Rio Tinto case when they were doing the TAR protocol. That was like bare-knuckle boxing to do that. It was widely reported. Everyone was talking about it. It was Da Silva Moore and then Rio Tinto, and it was like, wow, this is the big deal.
Kelly: Those were the Judge Peck decisions in which he essentially made it black-letter law that TAR was acceptable to use for purposes of document production.
Aaron: It’s black-letter law, and it was like seismic, right? What’s interesting is that a lot of the tools, a lot of the metrics and measurements that you would use in some of these newer tools are very similar to what you would do in a TAR 2.0.
Kelly: We need a Rio Tinto for Gen AI? Is that what you’re saying?
Aaron: We do. But it’s interesting, though, that what we are using a lot of are those older tools. Things like Reveal bought NexLP, which had Story Engine. We get a lot of value out of that tool that was 10 years ahead of its time, 10 years ago. And so we’re still using those tools to find amazing stuff right away because of the natural language processing tool that’s behind it and all the different models that underlie it.
Stephanie: I’ve been saying for a while that one of my favorite side effects of this Gen AI boom that people don’t talk about enough is that it has awakened more interest in technology across legal in general that I think is long overdue. It started with, Oh, throw some AI on that, which that’s not how it should work. But then you don’t always need a Gen AI. Sometimes it’s overkill. Sometimes you don’t even need AI. Sometimes a little bit of automation is just great. You’ve probably had some of those tools for five years and you didn’t even know about it, but now suddenly you know about them because somebody told you you can’t get the billion dollar Gen AI tool. So I think it is slowly reintroducing, like you were saying, older technologies that people might have overlooked.
Kelly: I’m going to go back to the prompt issue because I know you’re going to talk about this. But some of these tools all require you to input prompts. The question is, are those prompts ultimately going to be discoverable? I think that’s going to be an issue we’re going to be exploring. You were talking a little bit when we chatted, Aaron, about Cecilia at Disco as a chatbot, and you’re putting prompts into Cecilia.
Aaron: I’m not sure they would describe it as a chatbot, because I don’t know what connotations others might have, but it’s a question and answer functionality where you’re able to get a sense of what’s in there, ask questions, get answers. I’m basing this on descriptions from people who’ve used it. I haven’t actually used it. But when you have non-lawyers in there asking questions, some of these questions may not even be legal. It may just be, what’s the earliest document in the database?
Kelly: But I think there’s a lot of arguments. There’s a lot of discussion. It’s very factual. If I’m asking that because I’m trying to create a timeline of documents for purposes of deposition preparation, then that’s going to be work product. Even if it’s a paralegal or a non-lawyer in some other role doing it, if they’re doing it under supervision of counsel and they were asked to do it for purposes, it’s probably still going to be attorney-client privileged.
Aaron: You ready for this? I think the best way that you’re going to… You’re going to do many prompts. You’re not going to want people to go through and review it. So the best way to do it is with generative AI looking for what seems like a legal prompt.
Kelly: Use Gen AI to tell me what the Gen AI prompts are.
Aaron: Which is not that different.
Stephanie: There are tools already that are starting to preserve prompts, and I think more and more of them will do it, and I’m sure they are- What are those tools?
Kelly: I don’t know those tools. What are those tools?
Stephanie: I don’t want to speak out of turn. I’m pretty sure Casepoint is doing it. I know there are functionalities built into Relativity. I mean, it came up at Relativity Fest. But yeah, so it is. If you think at its very, very core, discovery requests are data related to this case, right? Prompts are data. The more people prompt, the more and more… We have more data than ever. When you were talking about that gap between the non-tech people and IT, I keep hearing more and more anecdotes that these people, especially, particularly in ediscovery, broad generalization, the partners were never in these tools. They weren’t getting their hands dirty. Now AI is so fun and cool. They’re like, I want to play around. So they’re getting in and they’re doing the prompts and creating even more data and maybe least knowing how to put in a prompt or understanding what to do. And not thinking that anyone might ever look at these things again. And I do think that prompts will be discoverable. We will get to the place where people are requesting and they have to be produced. But I do think a vast majority of them will end up being privileged.
Kelly: We have the same problem of technology being ahead of policy. Because how many people are able to preserve prompts? Does the technology exist to allow you to store your prompts? How do you produce them? What’s the context? There’s so many issues associated with another source of ESI that has a variety of things to consider.
Aaron: There’s also something I think is really fascinating, and I haven’t made the video for it yet, but it’s coming. In a scenario where you’re paying per prompt, you’re going to be really careful about the prompts. But in a scenario where you’re having a conversation, you may not be as careful with your prompts. And one of the issues with that can be prompt priming. And here’s the example. It’s unbelievable that it happened to me, but it’s great. I said, who are the top five ediscovery influencers on LinkedIn, and it gave me a list of five people. I was not on it. I was like, I said, who’s Gates Dogfish? And it described Gates Dogfish. And then I copy and pasted the first prompt again. And I said, I just pasted Who are the top five ediscovery influencers on LinkedIn? Gates Dog Fish is number one in the new answer.
Kelly: Wow. I think when you play around more and more, you learn all of those kinds of idiosyncrasies. It’s gotten a little bit better, but there’s sentence recognition. You can determine what’s written by AI based on the way the sentence is structured. It has some words that it uses over and over again. I get emails from people, whether they’re asking for a job or want to work with us or whatever, and I’m like, and AI generated, click, delete. It’s really apparent.
Stephanie: Can I make one more comment on that? I think per prompt pricing is stupid, but I think it’s become a knowledge management issue. If you have perfected, especially really long prompts that get you results you want within your organization, you should be building a prompt library so that everyone can normally reuse that prompt to create the same template, same document, same whatever it is you’re creating. So you’re not reinventing the wheel. You’re not reiterating your prompt as many times and adding more and more stuff to it. So prompt libraries themselves, so whether or not it’s discoverable, you internally preserving your prompts is a really good idea.
Kelly: Yeah, I totally agree with that. So a couple of things. One, I’m going to ask you guys, where do we see AI going in 2025, and how do we get people to get into the mix? What are some ideas we could use for people to use ChatGPT and just start understanding the capabilities of Gen AI?
Stephanie: I mean, in terms of getting in the mix, if you’ve never done ChatGPT, they now made it free for everybody. Just play around with it. Not for legal work, very clearly. Do not put your client data in it. None of that. But do a like, I want to plan my… I need a vacation itinerary. Things like that are not going to give away state secrets, and hopefully you’ll check it over before you get on a plane. That was right. But just play around and see what it can do. The dumbest little things. The first thing I ever did was write a song in the style of Bob Dylan about whatever.
Kelly: We did that for our ediscovery jingle a couple of years ago. We did. We used it to rewrite the lyrics to Christmas songs.
Stephanie: If it gives you something and it’s not right, try to iterate on your prompt and see if it gets better, see what gets you better or worse results. Try to give it as persona. That’s a huge one where I heard someone talking. There was a whole speech given by the French President about their energy policy, and there was one throwaway line that was like, We’re not going to be doing nuclear. Everything else was about it. And if you were just summarize this, they didn’t even mention that point. But then they modified it saying, You are an energy reporter. What are the key points about the energy policy? And it was at the top because it directed it what it cared about. So you can give it a persona. You can tell it how it should be looking through, especially if it’s a huge volume of data.
Kelly: You’re a marketing professional. Do this. You’re a salesperson. Do this. Yeah, all of those things. Some other ideas you can do, workouts. I think we talked about training sessions for kids for sports. Aaron, you put me on to the Coursera course, which is an intro to ChatGPT from Vanderbilt. It’s free. I signed up for it. You can just Google ChatGPT Vanderbilt, and it will come up. Cat Moon does a ton of really amazing stuff at Vanderbilt. She’s so great at her contributions. Just keep asking you questions. And Aaron, you mentioned this, ask me questions until you know enough, and then ask me what’s the next thing. Tell us a little bit about it.
Aaron: That’s my favorite. So I tell it you are a movie producer, and I am a screenwriter. I have an idea for a Wall Street thriller where XYZ. And then ask me questions about the setting, the characters, the ending, the conflicts. And I give it a whole list. Ask me questions about these things until you know enough to create an outline. Ask me one question at a time, ask me your first question. That way, you don’t get those long lists of answers that it can do. And I say, Don’t start writing the outline until you know enough. And before you write the outline, ask me whether I want you to write the outline. It will listen, and it does that. My favorite thing to do with it is if it asked me a question I don’t know the answer to, I say, What’s the best answer I could give to this question, and I have it answer its own question, and I would say eight times out of ten, it’s really a fantastic answer. Then I just adopt that as my answer because I agree with that, or it’ll be counterintuitive, and it’ll be something I wouldn’t have thought of because I’m so inside my own lane. That’s my takeaway for ChatGPT is if it asks you a question you don’t really answer, ask it to answer its own question.
Kelly: I think just get in the mix. I think as far as the case law goes, just because we’re on the case of the week, generative AI creeped into our case law in 2024. We added an issue tag for it to the database. There are five cases. It’s talking about Gen AI as opposed to being discovery about aspects of Gen AI. We haven’t really gotten discovery decisions. We’ve got a formal opinion from the American Bar Association on generative AI, the New York Times versus Microsoft and the number of companies that are suing Open AI over scraping the data. A few things on generative AI, if you’re a eDiscovery Assistant user, you can use that issue tag to be able to search for some of that case law. But I expect in 2025 that that will continue to grow.
We have a few minutes left and I want to turn a little bit to mobile device discovery, but do you have a couple of predictions or do we want to save those to the end? Or you’ve got predictions on Gen AI?
Aaron: Well, I’ve got one prediction on AI real quick. And I think that AI security is going to be massive in 2025. I just had a conversation with somebody at a holiday party where they were talking about how users were exfiltrating data by putting it, either copying and pasting it into ChatGPT or uploading the file into it and then accessing it from outside the environment. And so while it’s locked down, they can’t plug in a USB, they can’t drop it into Dropbox. They could copy and paste the information into ChatGPT and then get it on the outside. I think this type of exfiltration is going to burst onto the scene, and then so whoever can solve for it is going to be very happy.
Kelly: I mean, there’s always going to be bad actors, right? There’s always going to be people looking to game the system.
Aaron: But you don’t have to be a bad actor, right? You could just be taking source code. You could be taking source code and saying, Help me figure out why I’m getting an error on this.
Kelly: Yeah, but if you do that you’re essentially breaking privilege or whatever protection confidentiality you have over that information because you’re providing it to what is essentially a publicly available database. It takes all that data and it retrains on it.
Stephanie: It famously happened to Samsung. Somebody leaked their proprietary source code, not intentionally.
Kelly: There’s a lot of dangerous stuff. I totally agree with you. Stephanie, you got something on Gen AI prediction?
Stephanie: Agents, agentic AI. We don’t even need to get down the rabbit hole of what it is yet, but you might have heard it. It’s the biggest buzz, I would say, biggest buzzword in Gen AI going into 2025, and we will hear a lot about it. And like most buzzwords, some of it will just be hype, and some of it will be some really cool things. Be on the look out for that.
Kelly: I love agents. The difference between an agent and an LLM — an LLM, a large language model, has a whole bunch of data associated with it that you’re targeting with a prompt. An agent is something that you’re asking inquiries to take action vs. prompting an LLM. An agent is something that actually physically goes to the internet and does work for you. We use agents to do research. We want to know and understand what the company’s litigation portfolio is here and we break it down into a variety of things, go out to the internet and tell us what you can find, and it’ll come back. Or we want research on a particular person or whatever information you might need for purposes of a piece of litigation or marketing or sales. There’s so many ways to leverage agents, and they’re not hard to build.
Stephanie: Yes, and it doesn’t necessarily have to be the internet. It could be your entire system. This is the answer I want. This is the task you are supposed to solve. I’m not going to dictate every step of it for you. They intuit on their own and figure out their own way to get the right answer without as clear direction as just… Like with a chatbot, you’re like, ask a question, and they have a script somewhere where they’re reading from.
Kelly: We actually built one that allowed us to go through an entire litigation folder and create a series of events that we needed to from an old file, but it was all electronic. It’s really interesting stuff, but that allows you to build agents to stay inside your environment, so you’re not giving that information to anyone else.
Mobile Device Discovery
Kelly: I want to spend the last few minutes we have on what is, I think, the other biggest issue in 2024, and that is mobile device discovery. Almost every case that we are involved in as discovery counsel has mobile device issues. What’s come up, mostly in the case law this year, and we’ve done two or three decisions at minimum, probably a lot more than that, on failure to preserve text messages. There’s two cases in particular that I would encourage everybody listening to look at regarding text messages.
That’s the Maziar v. City of Atlanta, it was an employment discrimination case. They didn’t preserve the supervisor’s phone for more than a year, and they lost three text messages. The failure to preserve those three text messages meant that the court denied the summary judgment motion and required the city to pay the cost to the plaintiffs of the sanctions motion and the summary judgment motion. Probably more than a $200,000 bill for failing to preserve one cell phone. There are tools out there you can use. You can get a Cellebrite collection, which preserves the whole phone, and Aaron, I’ll come to you on that topic. Or you can use something like a ModeOne, which allows you to do a completely remote collection very cost-effectively. I mean, if you’re really cost-conscious, and you have the physical phone that you’re able to collect, you can even use something as inexpensive as iMazing. You just have to be able to authenticate the messages themselves. I mean, I may think it’s a desktop tool. It costs less than $100.
The other case that we covered on Case of the Week was the Safelite case in which the defendant left Safelite, went to another company in violation of his contract, according to the allegations. His lawyer told him to preserve all of his information. Eight months into discovery, he says, oh, the settings on my iPhone don’t keep text messages after 30 days. The counsel were on the fact that those messages were failure to preserve. But there was a step further that I felt like they could have gone because that is a setting that you have to physically change on your iPhone to set text messages to delete after 30 days. The default is to keep them forever. There’s a file on the phone that tracks every change that’s made to the settings of that phone. So They could have done a forensic examination of that phone, found out when he changed those settings. In that particular case, the reason that I suggest that is because the court found that there was not intent under Rule 37(e)(2), but there was prejudice under Rule 37(e)(1). They got an adverse inference instruction, but if they had been able to show that he’d made that settings change after the duty to preserve arose, which it seemed pretty likely they had just between us, then they would have been able to show intent. I think that would have been sufficient for intent under 37(e)(2), and they could have gotten a default judgment as opposed to an adverse inference instruction. The reason I think that’s important, and this is something that Judge Peck raised to me probably a year ago, was that adverse inference instructions may be great, but how many cases actually get to trial? If it doesn’t go to trial, you’ve got no sanction. It may be a settlement tool. Otherwise. But so in terms of mobile devices, Stephanie, I want to go to you. What are you in your conversations and the knowledge management and everything? What are people talking about?
Stephanie: I mean, I think it’s just because knowledge management comes from the other side for your own benefit, not necessarily for you thinking about litigation. But it’s just like, to me, the whole mobile device issue is, again, the realities of current technology. We all use mobile devices. We use multiple. And it’s not a new thing. I remember someone took a picture of me in 2007 when I was at the law firm, and I had an iPhone and a Blackberry and a bar napkin. I was doing work, and they jokingly posted it, and it was 0.6 hours. We’ve been using this stuff for so long. So again, and we talked about it earlier, if you have the abilities to retain this stuff and you’re not doing it, it’s not very long until that’s going to be hugely problematic.
Kelly: It’s already hugely problematic. So go out and get the tools, figure out what you need to do. You’re going to have to preserve mobile phones. So you’re out what you need to do.
Stephanie: I mean, in terms of actual litigation sanctions. We’re still doing that part. But it’s just to me, it’s a no brainer. I know you might think you’re being cute or whatever. This is me being totally blunt. You might think you’re being cute with all these technologies and being like, oh, we don’t have to do that yet. They’re just going to do this other thing. Just do it.
Kelly: I think we’re past the days of being cute. Yeah. Aaron, what are you seeing at Trustpoint? What are clients using, doing?
Aaron: Well, so I’ll show you just a quick meme here. I don’t have a lot of collection memes, but I have one on this one.
Kelly: Next Gen forensic collection. That’s a nice one.
Aaron: And it reminds me, do you remember what was the… I forget the name of it. I’m forgetting the name of the tool, but it was a pre-processing ECA tool that I think was purchased by Symantec. But they had a video where there was collecting custodians, and they were moving a wand around somebody’s head, and you could see the brain progress bar filling up.
Kelly: Like when Dumbledore has to pull the thoughts out of his head.
Aaron: Yes, into the pensive.
Kelly: It’s just to carry forth our Harry Potter references.
Stephanie: Harry Potter and Star Wars is strong.
Kelly: I know. Well, they are. I mean, they really are everything.
Aaron: And there’s so many of them. To answer your question, I know that targeted collections… How do I say this? Where do I start? We’ve done targeted collections and then had to go back to the source, and sometimes the phone has changed. Sometimes they’ve lost the phone, or they’ve upgraded the phone, or there are a lot of reasons that the phone might have changed, and then it’s problematic. And it would have been great if you had preserved the whole phone. But is that over-collection? Does the client actually want you to have their entire phone.
Kelly: That’s the biggest issue, I think, is the privacy of the client.
Aaron: Exactly. And so what the forensic partners that we have, what we do with them is it’s like, look, this is going on to a physically encrypted drive, and it’s going to be inside of an encrypted container. So this is just an insurance policy. There are so many cases of all these sanctions for spoliation. Why not just pay $250 plus the cost of the drive, have it right there, just in case. And we’re only going to do targeted searches. We’re only going to pull out things that are responsive to the date range, the search terms, etc. And we’ll work with those. And if it turns out that we have to go back, instead of inconveniencing the client, instead of finding out bad news, we have the insurance policy, the insurance policy image. We can go back to this has been sitting secure in our lab. It’s like double encrypted.
Kelly: But it really depends on your case. In both of the cases that I was talking about, the Safelite and the Maziar case, all they needed was a few text messages. If you know what you need from a phone, go and preserve just that. Use a targeted, remote if you can, easy collection that’s going to get you what you need and a backup report to be able to say when you did the collection and how you did it, and then document what reasonable efforts you took. If you learn six months later that you need new data and that’s gone off the phone, you’ve done what you’re supposed to do, which is take reasonable efforts. You did the preservation that was available at the time that it needed to be. There was no evidence that, say, a photo was going to be an issue until it came up way later.
Aaron: The documentation is key.
Kelly: The documentation is key, but early preservation is so key. I think the point that I want everyone to take away is there’s more case law now on possession, custody, or control being almost a non-issue. If your entities are encouraging or know that your employees are using their personal devices to communicate about work-related activities that are relevant or responsive in a particular litigation context, you are going to have to preserve and provide that information. Yes, we have outlier decisions on that scenario. It really does depend on the judge, and it’s not clear. It’s not black-letter law. But it’s really important, and I would just say this, it’s an insurance policy. It’s a really cheap insurance policy. I mean, for the city of Atlanta, how do you explain to taxpayers that you lost 200 grand when you could have even Cellebrited that phone for 1,500 bucks, right? I mean, that would have been overkill for three text messages.
Predictions
Kelly : All right, let’s do some final predictions. Stephanie, I’m going to let you start. Give me a couple of predictions for 2025.
Stephanie: Well, I already said agents, that’s really up there. Preserving prompts, it’s going to be a thing. Absolutely that. And then more generally, I just think we’re still going to see a lot more consolidation and funding in legal tech. I mean, the funding numbers have been through the roof. So much consolidation, merger acquisition activity, not necessarily where I expected it, but all of that is still going to be moving a lot. So the vendor you use today might not be here a year from now. That might not be a bad thing. That might have been bought.
Kelly: Fantastic. Aaron, what about you? Predictions for 2025?
Aaron: I think the AI exfiltration is going to be a big one. I think we’re going to start seeing case law about hyperlinked files outside of the communication context. Those are the two big ones.
Kelly: We don’t talk much about text messages or everything we share via WhatsApp or anything that we use. Signal is a huge one, which you can’t collect directly from a device. That’s a great point.
Aaron: That’s right. Something that got talked about for the first time that I have noticed at Georgetown was the role of private equity in ediscovery. I know that it’s there, and everyone knows that it’s there. But what no one is really talking about is the effect that it’s having. What does it mean for the people who are working at companies that are PE-backed? What does it mean for conferences that have all this money and new people showing up from PE firms? I think that the conversation around private equity in ediscovery is going to rise to the surface, and people are going to actually start talking about what it means, how to do it right, what’s the wrong way to do it, and also good outcomes versus bad outcomes.
Conclusion
Kelly: Those are excellent. All right, we managed to stuff as much as we possibly could into what became an hour and 15 minutes. So thank you to all of you who have stuck with us till the end. I think we probably still could have gone on much longer.
A huge thank you to Stephanie and Aaron for joining me today on this anniversary episode. Going into 2025, we’ll be continuing our Case of the Week series and adding additional segments on the podcast, which will be rebranded to Life and Law with Kelly Twigger. If you want to follow our additional content, please find our podcast on your favorite podcast platform and subscribe as well. Leave us a review. That helps us get to a wider audience so we can bring more folks into the fold on how to leverage the power of ESI.
Thanks very much to you both. Happy holidays to everyone, and we look forward to seeing you in 2025. Bye, everyone. Thanks.
