E290 - Doug Austin, Editor, eDiscovery Today
[00:00] Debbie Reynolds: The personal views expressed by our podcast guests are their own and are not legal advice or official statements by their organizations.
[00:11] Hello, my name is Debbie Reynolds. They call me the Data Diva. This is the Data Diva Talks Privacy podcast where we discuss data privacy issues with industry leaders around the world with information that businesses need to know.
[00:24] Now.
[00:26] I have a very special guest on the show, Doug Austin. He is the editor of eDiscovery Today.
[00:34] Welcome.
[00:35] Doug Austin: Thanks, Debbie. Thanks for having me and looking forward to having a fun discussion today.
[00:40] Debbie Reynolds: Yeah, it's almost not fair for me to just have you on the show and introduce you in such a way. We've known each other for like a billion years.
[00:48] I've had a very long career in technology and you knew me when I was doing a lot more advising with legal and law firms and companies to do legal tech and stuff like that.
[01:00] But I've always adored your work. You're incredibly smart. You bridge a lot of the technology and the legal side of things. And I think.
[01:10] Well, first of all, I want you to,
[01:12] to describe to listeners what eDiscovery is.
[01:16] Very, very niche industry and legal that people didn't really know and understand. But then it started because privacy has blown up in a number of years, a lot of that connection is starting to show more.
[01:30] But I want your thoughts there.
[01:32] Doug Austin: Yeah, absolutely.
[01:33] Well, yeah, eDiscovery, I think has initially started as somewhat of a niche industry. I think it's broadened quite a bit.
[01:40] And so with my blog titled eDiscovery Today,
[01:44] it's eDiscovery centric end stories I cover have a relationship in some manner to eDiscovery, which is electronic discovery. And basically that term originated from the legal discovery process associated with litigation.
[02:00] There's a discovery process of getting evidence from each party, from third parties,
[02:05] experts, things of that nature.
[02:07] That process for really probably more than decades, maybe even centuries at this point has been called discovery in the past 30 plus years where more of the evidence has become electronic oriented,
[02:20] electronically stored information,
[02:22] electronic discovery or e discovery has become a big industry.
[02:27] And one of the things that I cover with regards to eDiscovery is the fact that eDiscovery isn't just about litigation anymore. It's about investigations. Those could be internal or regulatory investigations.
[02:40] It's about audits,
[02:41] it's about data privacy,
[02:44] things like dsars and things like that. Because that's a. There's a discovery workflow associated with that. It's about incident response where there's a discovery workflow associated with finding out what personal information has been exposed.
[02:58] And, and notifying those individuals of their personal data being exposed and having a process to that. It's even associated with second requests for mergers and acquisitions.
[03:10] So Discovery and eDiscovery has really broadened. And my blog really touches on anything that's really related to eDiscovery. So that could be cybersecurity trends, could be data privacy trends,
[03:23] it could be information and data governance.
[03:26] And of course,
[03:27] the elephant in the room these days is AI.
[03:29] And a lot of that relates to everything we do and it certainly relates to E discovery in a big way.
[03:36] Debbie Reynolds: That's a great primer for anyone who doesn't understand that. I remember.
[03:41] So I write a chapter in a book for Thomson Reuters and have been doing it, I don't know, 15 years or so. And I was the first person to introduce data privacy into that book because I'm like, these things are related and we need to be talking about them together.
[03:58] Some of the parallels I see in my view, and I want your thoughts. And I think it's coming even more highlighted now that the AI is in the room. Right.
[04:07] And that's about authenticating things,
[04:11] finding data, using it in ways that are legal, not breaking the law. Right. Because we know it's super easy to do that with electronic data. And so as opposed to boxes that are behind a door, you know, you have data that flows like water through data systems.
[04:28] And so that's how I got interested in it because a lot of people who knew me from some of my enterprise work started calling me about privacy because they're like, well, we know that you know how to do this.
[04:41] We know you know how to move data from China to the US without like breaking laws. So this is something that I thought was really interesting. But I love your work in eDiscovery today.
[04:52] I highly recommend people look at it because it is very cross sectional in terms of the technologies that you talk about and a lot of the cases that we're starting to hear a lot about.
[05:03] But I want your thoughts.
[05:05] Doug Austin: Yeah, well, first of all, you know, data people talk about AI so much, but data, in my opinion, equally important and really in many ways more important because data is a driver of AI.
[05:17] It's data drives all the AI models that we're all using in work and in our personal lives today.
[05:24] It's driving organizations, you know, in terms of their ability to get the information they need to, you know, to make decisions,
[05:32] to manage situations like litigation and investigations and what have you.
[05:38] So the fact that I'm speaking with the data diva is definitely an honor for me because data is really what ties all these, you know, disciplines together.
[05:49] And you know, I think that when it comes to authentication of data, obviously one of the big AI related topics is deepfakes. And you know,
[05:59] there's certainly more AI generated acknowledge, what we would call acknowledge AI generated evidence out there from all the, the AI chatbots like ChatGPT,
[06:11] from things like Copilot, artifacts that people need to be thinking of and certainly even like the AI recorders that are on meetings. I did a webinar yesterday and somebody's AI notetaker hopped onto the meeting first thing in the chat.
[06:26] It was a zoom webinar and that's pretty common these days. So that's all AI generated evidence or information that could be evidence in investig litigation and investigations.
[06:39] And I definitely think that it's just creating more data for us to figure out a way to manage.
[06:45] Debbie Reynolds: Very true. Actually, I saw an article in 404 Media about some company taking people's public links to meetings. And I guess they didn't explain how they did it, but I assume the way they do it is they invite a note taker to a meeting and maybe the meeting is crowded,
[07:05] people don't notice. But then this company is taking that data and actually creating webinars. Like I guess they put it in notebook LLM and create like a podcast and then they put it on a platform and they literally sell it to people.
[07:21] So there are people like, well, private meetings that they didn't know, that someone knew about and it was literally on this platform.
[07:28] Doug Austin: Yeah, well, and that's one of the things about,
[07:31] you know, a lot of people, it's kind of like, you know, you, you've been, we've all been given these toys to play with and not the instructions or rules without play how to play with them.
[07:42] And so we're, we're seeing people do things like loading company documents into a, into a public LLM like ChatGpt or Claude. We're seeing people do what you talked about, which is recording a meeting or you know, transcribing a meeting and doing note taking and not informing the people that they're doing that,
[08:01] which of course you should be doing.
[08:03] Certainly. I think in many meetings people are like, yes, and as long as it's used for this for note taking purposes,
[08:10] that's fine. That's one of the things I always try to do when I'm conducting a meeting where I plan to use an AI. Note takers say, hey, I'm just recording this for note taking purposes.
[08:19] Won't go beyond that. Are you okay with it? And if they say yes, then I kind of move ahead.
[08:25] I think that's one of the things that people don't think enough about. They're thinking about, gee, these things can, these tools can really do a lot of great things,
[08:33] but they're not thinking about the data privacy implications or the potential exposure of confidential information that could occur if you're not using these tools properly.
[08:45] Debbie Reynolds: One of the things that concerned me, I said I was in a meeting once and a person had a note taker and I guess their note taker, they said they had a set up where everyone in the meeting got a copy of the transcript or the summary after the meeting.
[09:01] But this is one of these meetings, like typical corporate meeting, where one part of it was a general meeting and then the other people dropped off and then it was a more confidential meeting,
[09:13] but the note taker sent all the notes to everybody, even the people who weren't supposed to be in that second part of the meeting. And I think a lot of companies don't think about the implications of that.
[09:24] Doug Austin: Yeah, I mean, they're really.
[09:26] There are so many kind of things that organizations need to think about,
[09:30] and those are absolutely one of them. One of the considerations that I talked about on a webinar I conducted about AI yesterday is the whole idea of the, you know, opponent potentially taking your production and loading it into a public LLM.
[09:45] And I actually just covered a case last week, Jefferies vs. Harkrose Chemicals,
[09:51] where the parties disputed over amending a protective order. They already had a protective order that said that the no party will upload any confidential documents into a public LLM. But they were fighting over whether the plaintiffs could upload non confidential documents.
[10:08] And the court agreed that from a data privacy consideration, cyber consideration and what have you, there's no way that the plaintiffs could guarantee that data being uploaded to a public LLM like chatgpt couldn't get out there somehow and couldn't be necessarily used to train the model.
[10:27] So the court agreed to update the protective order to protect all of the productions. And I think that's an example of something you're going to see standard in litigation these days.
[10:37] Because I mean, while I think a lot of organizations might understand the ramifications of doing that, your typical pro se plaintiff may not. They may just get a production from the defendant and they say, well, gee, I want to understand these documents better, I'm going to load them up in the ChatGPT,
[10:54] they shouldn't be doing that. And protective orders are needed to avoid those types of things. So that's just one example of the types of protections we need now from how AI tools are being used.
[11:07] Debbie Reynolds: And I want to talk a little bit about a case. We're going to talk about a couple of cases, but one I wanted your thoughts on. I'm sure you're following very closely, and that's the New York Times vs OpenAI case, where the judge had said that OpenAI needed to retain all of their data because of their litigation that they had.
[11:26] And this is kind of a massive,
[11:28] huge thing that has implication not only on litigation as we see it, but also privacy,
[11:35] because it's like people say, well, I put up my own ChatGPT account and I said, I don't. Don't save this data. I'm like, but,
[11:45] you know, when you have a legal order that supersedes your setting within the tool. Right. And so we're seeing people in even other cases running now to OpenAI to get data from that company because they know that they're under this order.
[12:01] But what are your thoughts?
[12:04] Doug Austin: Yeah. When the court ordered OpenAI to keep so many of their chat logs, I mean, like, literally, you know, hundreds of millions of chat logs,
[12:11] that obviously they.
[12:13] They did, and I can't remember exactly how many months that was in effect. Ultimately, the parties agreed that they could suspend that preservation of such a large collection. And they've been, of course, arguing about the number of prompts and responses that they needed to produce.
[12:31] I think they ultimately agreed on something, some crazy number like 200 million,
[12:36] which was a compromise from what New York Times wanted. I think they wanted much more.
[12:41] But that, say,
[12:43] a logical example of how AI related evidence is becoming more important in litigation these days. And obviously there's considerations about how the AI companies are storing this data. You know, what they can long they need to store it when they need to provide, you know, produce.
[13:01] And when these companies are involved in so much litigation,
[13:05] especially copyright litigation,
[13:07] that I can imagine their data processes are a legal hold nightmare at the moment, because it's probably pretty hard to know exactly where you can establish your legal holds and what can be safely and defensively deleted versus what has to be held for various litigations you're involved in.
[13:25] So, to me,
[13:27] these are the cases that I think are helping to define just exactly what we can expect with regards to these AI companies, what they're doing, and,
[13:36] you know, what data they'll have to produce to defend their actions. I also think that we're beginning to see a Lot more cases involving AI related data that don't involve AI companies, but just organizations who might have some person who used ChatGPT to maybe get ideas of how to work out of a contract or something or things like that,
[13:58] where that's going to be one of the things that an organization may have to preserve and may have to produce.
[14:05] And I honestly think that AI generated data may become one of the biggest categories of modern data that we've seen yet. Because, you know, we already have mobile device data, we have collaboration, app, cloud,
[14:20] enterprise solution data. And now AI generated data is just in so many different forms that I'd expect it to become maybe one maybe the biggest of them all.
[14:31] Debbie Reynolds: You have a case, big case that came out about privilege that I want to talk about. But as you were talking, I was just thinking when people are interacting with a chatbot, they're like iterating, they're thinking, they're giving their thoughts.
[14:48] And in a way it could be considered if the person were a lawyer, it could be considered a work product.
[14:54] But if you're not a lawyer,
[14:55] you know, there's no work product doctrine that covers kind of a non, like a pro se litigant or different things like that. But that seems like a very interesting issue here, especially as people.
[15:08] Well, first of all, there are so many self represented people in the legal process. Unless you're a big company,
[15:15] a lot of litigation does involve people who are not represented by lawyers and they don't have the same protections. Right. And I think we're gonna start to see more, or we are already seeing more pro se litigants using AI in different ways.
[15:31] But what do you think?
[15:33] Doug Austin: Yeah, yeah. So a couple of things to unpack there. I'll start with the privilege stuff first.
[15:37] And certainly we've seen a couple of recent cases which really were like right there together. In fact, the oral ruling on one case happened the same day as the written ruling on the other case.
[15:49] The first case that I'll talk about is the is US V. Heppner. And in that case,
[15:56] basically what you had is a criminal defendant who created some documents using Claude and ultimately provided them to his attorneys.
[16:06] And the judge in that case examined both attorney client privilege and attorney work product,
[16:12] decided that it wasn't attorney client privilege because he was doing this on his own. He wasn't doing it at the direction of his attorneys before he provided that information to them.
[16:23] So it wasn't work product because of that, and it wasn't attorney client privilege because it wasn't a lawyer directed communication.
[16:32] So in that case, basically the court found that using an AI chatbot, no reasonable expectation of privacy,
[16:40] that the documents weren't protected by either a work product or attorney client. So the Senate's had to produce those documents.
[16:48] Interestingly enough, Claude has an opt in for training the model versus ChatGPT,
[16:55] which has an opt out for training the model.
[16:57] So technically,
[16:59] assuming most people wouldn't change their default if they're using Claude, their data isn't training the model. So it will be interesting if we get to a point with a case someday,
[17:09] whether the opt in or opt out will become important in determining whether there's a reasonable expectation of privacy.
[17:16] The other case is the Warner v. Gilbarco case and that you mentioned pro se parties. That was one where the plaintiff was a pro se party and she was using ChatGPT to create some documents.
[17:29] And in that case the court found that it was attorney work product because in essence she's acting as her own attorney. So the stuff she's creating there is work product and was and determined to be protected in that case.
[17:43] Another kind of of offshoot to the whole AI discussion and case filings and what have you is all these case filings with hallucinated case icons or fake information within, you know, real cases or what have you.
[17:58] And we're seeing a lot of, you know, I think Damian Charlatan's database were over 1200 cases now,
[18:05] 700 plus of which were just last year. And we're already, I think 300 plus or so in the end this year in early April.
[18:13] And a lot of people think, well, lawyers really need to get on the stick. But there's more pro se instances in that database than there are lawyer databases. I just checked a couple of days ago because I was talking to somebody about it.
[18:25] It was 700 plus pro se parties with FAA citations,
[18:30] 400 and some odd lawyers.
[18:32] So not a good number necessarily, but yet another ramification of how we're using these tools without really understanding, you know, fully how to use them and to checking the results and making sure that is valid.
[18:47] Debbie Reynolds: That's fascinating. And one of the things to me that fascinates me about the hallucination thing,
[18:55] especially with the lawyers,
[18:57] not the pro se litigants. And so to me it's different with a pro se litigant. First of all, legal research is expensive. Unless a case is like really famous or public,
[19:08] it may be hard for you to get access to it. Cause a lot of lawyers have access to paid research tools. But for the lawyers who have access to paid research tools or who have been litigating for 30 years.
[19:20] I'm like, for example, if you're an antitrust lawyer and you're on a case and you find 30 new cases you never heard of,
[19:33] it's probably a problem. Right? There probably aren't 30 new cases you hadn't heard about in the last 30 years. Right. So that should be a tip off right there that you need to really look more deeply there.
[19:45] Doug Austin: Yeah, absolutely. And I mean, I think with. Because I probably get sent one or two cases a week from people saying, can you believe this happened? Can you believe this happened?
[19:54] And I mean, the sanctions have been all over the board. They've gotten tougher. They've been, you know, they've included significant financial penalties. They've included reporting to the bar. They've included, especially if the lawyer, which in some cases lawyers really, they don't have responsibility for their failure to check their citations.
[20:12] So you, you, these more significant sanctions.
[20:15] And I think we, many of us in the industry remember almost three years ago now the motto of the Avianca case and how publicized that case was. I mean, covered not just within the legal sphere, but on CNN and New York Times and national publications.
[20:31] And you were thinking, okay, the word's out. People now need to know. No, they need to check this stuff. And they're not doing it. And even these significant sanctions aren't keeping them from doing it.
[20:41] And I honestly think it's.
[20:43] I wrote a blog post about it a while back that I think this is something that a phenomenon we refer to as automation bias,
[20:51] where we tend to trust an output from a computer or from an automated system, because it's an automated system, we think it's likely to be right. That's the whole phenomenon behind death by gps.
[21:03] People driving off a bridge because they're trusting their GPS and not realizing that, hey, the road actually doesn't exist here or driving into the ocean or something like that.
[21:13] So. So I think that's one of the challenges here. And unfortunately, while people keep thinking, well,
[21:20] when this sanction doesn't get people's attention, nothing will, well, I can't imagine any more significant sanctions than have been already issued to some of these lawyers. And yet we continue to see several of these a week.
[21:34] And I don't think that's going to change anytime soon.
[21:37] Debbie Reynolds: I think the AI can be beguiling,
[21:40] I guess,
[21:41] and it absolutely gives you the impression that it is confident in the thing that is said. And then you think, oh, wow, this is a shortcut to My research,
[21:53] because they gave me the exact statement from the case or whatever.
[21:56] And then if you go on the Internet or go into a search tool that you pay for and you look up that same case, it's like completely different.
[22:07] They may have gotten the case name right, but the case subject is totally wrong. The quote they said is not there. And so that's always concerning. But I think always my feeling, just being a technologist for so long,
[22:20] I think people want to have like an easy button.
[22:25] So that's what, that's what they desire.
[22:27] So when they see something that looks like, oh, this could be it, this could be that easy button that I always hope for, and they go for it and it's not really there,
[22:36] you're
[22:37] Doug Austin: preaching to the choir. I've been saying the exact same thing. People want the easy button. You know, electronic discovery. There's been a discipline that's existed for facilitating and automating a document review known as technology assisted review or predictive coding.
[22:51] It's been around for 15 plus years.
[22:54] It's a say, kind of. It's a process that involves statistics and what have you, but it's also a process can enable you to reduce the number of documents you actually have to reduce, can reduce costs and make you enable to get through a large review more quickly.
[23:09] But it's never really caught on to the extent people thought it would. And I think a lot of reason is because there's a, there's some diligence to it, there's a process to it and people want the, the easy button.
[23:20] You know, a couple of examples that I'll, I'll kind of mention just from a,
[23:24] that I've talked about and, and one I experienced,
[23:29] we've seen from time to time, like some of these,
[23:33] you know, Google examples like telling people to eat rocks or put glue on a pizza to make the cheese stick or stuff like that.
[23:40] But another one that came up was that you asked Google, is Lady Gaga only two days older than Ariana Grande?
[23:50] And Google would say yes.
[23:52] And then what? Google would say yes,
[23:54] but then proceed to list their birth dates as about five or six years apart.
[23:59] And you're thinking, well, why is it saying they're two days apart when their birth dates are five years plus apart? Well, there was an article where the two of them were rehearsing for a show and clowning around and Ariana Grande said jokingly to Lady Gaga, oh, you're like two days older than me or something like that.
[24:17] So apparently the model picked that up and ran with it. So those are the types of things where sometimes there's a little bit of foundation to that hallucination.
[24:26] Another one is one that I experience personally with anytime I get a new client.
[24:32] I asked ChatGPT to do a deep research on their website, just kind of give me information about their services, you know, things that the case studies they worked on, stuff like that.
[24:42] And this one of them came back with a real good description of a very involved case involving mobile devices in three different continents and whatnot. And I'm thinking, wow, this is a really good case study.
[24:55] I need to find that case case study on their site. Couldn't find it. So I went back to ChatGPT and asked about it and it came back and said actually I should have flagged that as a hypothetical.
[25:06] So it stated it as if it were real, but it wasn't real.
[25:10] So those are the types of things you have to kind of watch out with for with the technology.
[25:15] Debbie Reynolds: What are things that are concerning you most right now? What are you seeing this like just you're thinking, oh wow, this, I don't like the way this is going or,
[25:24] or this needs more attention or more focus.
[25:28] Doug Austin: Well, I think that one of the things that I'll kind of take it from a standpoint of and continue on that last vein of discussion is validation of the results from an ediscovery standpoint.
[25:43] One of the things that the best practices started with technology assisted review is validating the results and being able to say with confidence that this is the recall, this is the precision.
[25:56] We've gotten it to a level where we feel comfortable that, you know, most of the responsive information has been produced and very little non responsive information has been produced. And it's a process often iterative to get there.
[26:10] When it comes to using generative AI for document review,
[26:14] it's a. The same validation techniques can apply in terms of testing for recall precision. What have you, you. But generative AI has so many other uses and capabilities and there's really no validation kind of standards for things like confirming that document summaries are correct or confirming that the information it provides you when you're doing early case assessment and figuring out what your document collection has that's helpful and harmful to you is correct and things like that.
[26:43] So I think those are, you know, among the things that we're going to have to really solve to get the full capabilities out of this technology.
[26:53] And then of course the other thing that I think is concerning is obviously the impact on people's lives and jobs and things like that,
[27:02] but one of the things I think when it comes to that is AI doesn't replace jobs, it replace tasks.
[27:09] So the key being a person who can protect against that is to be, be good in as many tasks as possible and be able to be versatile in that. Okay, maybe AI is taking most of the document review and automating that.
[27:23] But I, but I abilities to do these other things, to provide the analysis, to kind of develop strategy, to do things like that that maybe AI won't be as good at and so forth.
[27:34] So I think those are the accuracy and the, and the considerations are the first and foremost. The last one is pricing. The technology is still pretty expensive when you're using it for things like eDiscovery,
[27:47] but that's pretty common with new technologies in eDiscovery. They start out high. You know, I remember, you know, a couple of decades ago when it cost $250 to process one gigabyte of data.
[27:59] Now, of course, it's free because the technology's evolved so much. I think that's just a natural progression that's going to happen with the AI models as well. That pricing will come down, it'll just take time.
[28:11] Debbie Reynolds: What is your thought about privacy and how privacy seems to be playing in more,
[28:17] more and more in these cases in ways that maybe people would not have ordinarily thought they would?
[28:24] Doug Austin: Yeah, I mean, I think that,
[28:27] I think that every one of the things that we deal with in eDiscovery is what we call the modern data challenge, because data is in so many different forms than it used to be from an eDiscovery standpoint.
[28:39] We used to deal with office files and email, and that was the large majority of the electronic evidence we dealt with in cases. But now we deal with mobile device data, we deal with chat, collaboration apps, we deal with enterprise systems, and we're dealing with AI generated data.
[28:56] And all of those have a different data privacy challenge to them in terms of how that data could be made, inadvertently made available.
[29:05] And of course, eDiscovery is typically a process where you're duplicating copies of data as you're moving it through the cycle. So you're exposing, you're giving greater chances to expose that data when you're doing so.
[29:18] So to me, I think that the more we get into more modern data sources and, you know, the, the types of data that we need to manage in organizations proliferates, the more difficult it is to secure that data from a data privacy perspective.
[29:35] And one of the things I've noticed is that Data privacy seems to be taking a backseat to the full speed ahead. Kind of thinking of, let's go forward with AI capabilities and what have you, even so much that it seems like even the EU is starting to back up on some of the GDPR requirements because they don't want to be left behind in the AI rush.
[29:57] So I definitely have concerns about that, and I'm hoping that. That at some point we will kind of wake up and realize,
[30:05] yes, it's important to advance the technology and really maximize the capabilities of what we can do with this technology.
[30:13] But we can't do so at the expense of exposing organization or personal data and making it just so easy for people to get to that data.
[30:23] Debbie Reynolds: I think one of the things that concerns me a lot about AI, first of all,
[30:28] the technology is advancing so rapidly and some of the problems that can happen to people,
[30:36] to me,
[30:38] there are things that can happen with AI for which there is no adequate redress, in my view.
[30:44] So I'll give you an example. And I'm sure you've heard of the story about the woman who was jailed. She was sent to a different state she had never been to.
[30:53] She was put in jail for six months.
[30:55] They said she committed some crime. First of all, she had never been in the states. Okay?
[30:59] Doug Austin: Right.
[30:59] Debbie Reynolds: They took her to the state. She lost her house, she lost her job, she lost her dog. You know what I'm saying? And it's like the state was basically like, oops, we're sorry.
[31:09] And I'm like, so how can you. What redress is there really for this person? You know? And this is what concerns me with AI.
[31:16] Doug Austin: She lived in Tennessee and the crime happened in North Dakota. She'd never been there.
[31:21] And I think she was four months in jail in Tennessee before she got moved to North Dakota. During those first few months, she wasn't even able to get an attorney.
[31:29] And finally, when she was able to get an attorney who could prove she was in Tennessee during the crime, she was able to get out. And. Yeah, facial recognition.
[31:37] I just don't understand. I mean, it just seems so logical to say, okay, you want to use facial recognition to identify potential suspects, okay. But then you have to have other data that shows they were actually there.
[31:52] You have to prove it. You have to be. Because we know this technology can make mistakes.
[31:57] So you have to have some way of really putting them there beyond just the facial recognition.
[32:03] And it really wouldn't have taken much of a check at all by law enforcement to have figured out, well, this lady clearly was in Tennessee. Never even been to North Dakota and clearly wasn't the person that the facial recognition algorithm said she was.
[32:18] So yeah, that's. And really when we're talking about some of the stakes,
[32:22] stakes aren't much higher than situations like that.
[32:26] So yeah, that's another example of trusting the technology too much and abandoning what should be sound law enforcement procedures of gathering evidence to put the person at the scene of the crime, which is what people used to do when there wasn't technology to do that.
[32:43] That was what they were expected to do. They didn't always do that. But now relying too much on the technology is making them not do the job that they need to do to confirm that that person is the actual suspect they should be looking for.
[32:58] Debbie Reynolds: In my view, technology has always been about helping someone do heavy lifting.
[33:04] But in my view, figuring out that someone has never been in a state is not super. First of all, it takes no technology to do that.
[33:14] So let's start there.
[33:16] There are some good old fashioned police work that needs to happen here before you kind of do stuff.
[33:21] Also from an evidence perspective,
[33:24] like, like evidence, like facial recognition,
[33:27] in my view those things should be like corroborating stuff that you already have, not being like the lead piece of evidence that you have. I don't know. That's my.
[33:37] Doug Austin: Well,
[33:38] I mean I, I think it's okay to consider using the technology to identify potential suspects, but I emphasize the word potential.
[33:46] That can't be the sole.
[33:48] The evidence you have.
[33:50] You've got to have other evidence that proves they were at, you know, they were there and had the scene of the crime and potentially committed the crime.
[33:57] We've seen too many instances where. I know you've written about it, I've written about it. Where people have. Law enforcement has just gone by the facial recognition and ran with it and ultimately that the, no, no other evidence really existed to prove that that person committed that crime.
[34:15] So yeah, I mean, I understand using the technology to identify potential suspect, but then have still gathered the evidence that actually proves that they committed the crime. And I think there have been plenty of cases where law enforcement has skipped those steps and hasn't done so.
[34:32] Debbie Reynolds: We're starting to see people use more deep fake technology sort of in cases as evidence.
[34:38] I think a couple years ago there was a case where I guess it was a murder case or something and they said they had some grainy video footage of what parts of the,
[34:50] and what they thought happened. And whoever had the video, they had it like digitally enhanced in some way or they had altered it in some way.
[34:58] And so there was a big dispute about this. It's like, well, why don't you give us the raw video as it is?
[35:05] Because they literally took it to like a studio or something and they tried to make it more dramatic,
[35:11] they altered it in some way. And to me that's like the opposite of what you're supposed to be doing with evidence.
[35:16] But I don't know. What do you think?
[35:18] Doug Austin: Well, you know, I,
[35:20] I mean, there we have. We've certainly used technology to try to make video clearer. I mean, I'm a big Through Crime fan. I've watched Forensic Files. I've probably seen many of the Forensic Files episodes multiple times.
[35:32] They've had some episodes where they've sent like surveillance video to NASA and they've used technology to clean it up and make it a little clearer to be able to kind of figure out maybe who the suspect is and what have you.
[35:44] And I think, I think there can be some value to that, but I also think there has to be some understanding of what the technology could do that could maybe take it a little bit beyond what the video looks like.
[35:58] And yeah, I think there's.
[36:00] That's one where it's kind of hard to say. I mean, I don't think I'll use the lawyer, Lawyerly answer of it depends because I don't think you can always say no, you should never use technology to enhance video or you should always use technology.
[36:12] I think it kind of depends on the situation and what you can do and how you can describe the process.
[36:19] Because it can't be a black box. It has to be a describable process where you can say this is what we did to enhance this video to enable us to see clear what happened here.
[36:28] I remember one of the cases that covered, I think it was a Forensic Files case where they were able to enhance a video well enough to see it was basically a lady who was abducted and killed.
[36:40] When they enhanced the video enough, they were able to see that it was her ex husband's car that she wound up going in and he was the one who ultimately killed her.
[36:49] They wouldn't be able, wouldn't have been able to do that without enhancing that video. And then they were able to pair that with other evidence to ultimately convict him.
[36:57] So I definitely think there's a place for using the technology, but I definitely think it has to be explainable and understandable what exactly the technology did with evidence like that.
[37:10] Debbie Reynolds: It's a very interesting time to see the way technology is being used. I Feel like. Like we're at a time.
[37:18] I'm getting a feeling right now the way we're using AI and the things that we're doing and the evolution of artificial intelligence.
[37:27] The same feeling I got when the commercial Internet started, in a way.
[37:31] Right. So it was a lot of excitement, exuberance, a lot of things that we wanted to do. We were really fascinated about what we wanted to do. And so that's kind of the feeling that I get and the vibe that I get.
[37:43] I remember back in those days and olden days,
[37:47] you know, I, I knew lawyers who, like, they would never touch a computer. Like they said, like, they just thought it was like a fad or they thought our computers were for secretaries, they weren't for lawyers.
[37:58] And I've seen, I've been through that whole cycle, so I've seen that whole thing go. And I think we're having some of the same arguments with AI, but I think because the technology is advancing so rapidly that those things are happening.
[38:11] We're seeing in real time the cause and effect of the. The uses of AI. But what do you. You think?
[38:18] Doug Austin: Yeah, absolutely. I mean, you know, what's the. I don't think it originated with Spider man, but with great power comes great responsibility.
[38:26] You know, I think we forget that, you know, too, too often.
[38:29] But, yeah, I mean, you have to embrace the fact that progress happens and it's important to be part of the solution. Not kind of put your head in the stand and say, I'm going to avoid this technology because there's still, you know, unanswered questions about it.
[38:43] But instead, I think being, you know, having a way of being able to,
[38:47] you know, use the technology effectively,
[38:50] verify the results,
[38:51] you know, be careful with data when using the technology,
[38:55] especially your own data or someone else's data or your organization's or customer's data.
[39:02] I think there's best practices that involve all of this. It's not a staples, easy button. There's gotta be a process and a workflow to using these tools the right way,
[39:12] and you can do so and progress and be able to do more things. And I think even lawyers are understanding that they're embracing the tools. It's almost kind of like the.
[39:23] Well, now we wanted lawyers to embrace the technology, and now with all these hallucinated cases, we're saying, okay, well, now here's what we get. But I think it's understanding and learning.
[39:33] There's a right way and wrong way to use these tools, and you've got to verify and validate the results always
[39:38] Debbie Reynolds: ways think that's true wherever we're in the world, according to you, Doug, and we did everything you said. What would be your wish for privacy anywhere in the world? Whether that be regulation,
[39:51] human behavior, or technology.
[39:56] Doug Austin: Gosh, there's narrowing it down to one might be tough. I would like to see at some point us as a society get, get together, agree on stronger protections for data privacy in place.
[40:09] Because it does seem like it, you know, even before the AI boom of the past few years,
[40:15] social media companies and data privacy have been a big challenge for, you know, the past couple of decades.
[40:22] So to me,
[40:24] one of the things that seems to not be in the, you know, in the horizon at all is like a US national data privacy policy.
[40:34] Other countries have done it, we've had a few states to it here, but certainly not the nation. Doesn't look like it's happening anytime soon.
[40:42] So I would really want to see us start there and get more rigid with what companies can and can't do with people's data.
[40:50] And I think that the only potential way that organizations might be forced to change their ways is litigation. We're seeing some cases against social media companies.
[41:03] Meta got hit with a couple of judgments last week.
[41:06] I think that's ultimately going to be the way that maybe will force companies to be more rigid with how they manage, with how they handle people's data than they have been up till now.
[41:20] Debbie Reynolds: Well, I support that wish.
[41:22] I think also in the,
[41:24] I don't know, I guess in AI and privacy, definitely privacy. So privacy, because data is so fundamental to these systems,
[41:34] that's the foundation, I think, that we're missing.
[41:37] But also one of the things that I find concerning, and I think you have brought it up in two of the different cases that you have brought up about the privilege situation where one judge said you waived privilege and the other said they didn't.
[41:53] It's like a work product.
[41:55] It's like without any rules or some of these things, it's just up to the judge. Right. And so what that creates is a situation where there really is no standard.
[42:07] It's just you're trying to hold on to a precedent at the time and they can go in any different direction.
[42:14] Doug Austin: And unfortunately, hey, that's the litigation world we live in. Many of these things are not bound by federal rules or state rules or what have you. They're bound by precedents and case law.
[42:26] That's why, you know, covering case law, I usually cover between 60 and 70, 70 some odd cases a year. You know, that's one of the things I really, you know, try to emphasize as part of my coverage of an ediscovery today,
[42:40] that case law really determines how we manage ediscovery. And it also has ramifications in these other areas and data privacy and protection and the proper use of AI and what you can and can't do with it.
[42:53] And you just gotta hope that enough judges will be mindful of those of the right things to kind of say, hey, you know,
[43:00] these are guidelines that make sen.
[43:02] You know, anything out over here is out of bounds.
[43:05] Debbie Reynolds: Well, I very much support your work. I love the things that you do. I love to see in the morning your posts pop up. Cause there's always a really deep dive into particular issues, especially around cases and what they mean.
[43:18] So. Yeah, well, let people know. How should they reach out to you? They want to get involved in your work.
[43:24] Doug Austin: Well, first of all, back at you, Debbie. I love your work as well and definitely am excited and honored to be on the Data Divas podcast.
[43:33] So, yeah, yes, my blog is ediscovery today. That's ediscoverytoday.com, all one word, no dashes. I will just mention that it's a daily blog covering all those topics. eDiscovery, data to privacy,
[43:47] CyberSecurity, InfoGov and AI.
[43:49] Actually,
[43:50] usually two to four posts a day most days, most business days.
[43:55] And I also will say that if you subscribe, we have certain reports that we produce throughout the year and subscribers get those for free. So. So that can be something else to check out.
[44:07] Debbie Reynolds: Thank you so much. I appreciate your friendship and I appreciate your work so much. So thank you so much for being here.
[44:15] Doug Austin: Thank you, Debbie. And hopefully we'll see each other in person sometime soon.
[44:19] Debbie Reynolds: That would be great. All right, talk to you soon. Thank you.
[44:22] Doug Austin: Thank you. Bye.
