Latest / Coopers' Code / ChatGPT and Attorney-Client Privilege, Zuckerberg on Trial | This Month in Legal AI | Coopers' Code

Transcript

ChatGPT and Attorney-Client Privilege, Zuckerberg on Trial | This Month in Legal AI | Coopers' Code · Coopers' Code

• 0:00
  AI has gotten more legal friendly, lawyer friendly.
• 0:03
  I'm finding that we're we're engaging with more educated
• 0:07
  consumers. What I mean by that is in the
• 0:10
  past in the personal injury world, often times somebody was
• 0:13
  coming to us with, with 0 experience and it and one of the
• 0:17
  nice things. Now it's it's both nice and a
• 0:20
  problem is that they're much more educated that they've done
• 0:24
  some digging on AI and have some sense they may have even drafted
• 0:28
  a demand package that they've sent off to the insurance
• 0:30
  company themselves before they've come to us.
• 0:32
  The downside is the education that they've gotten is is is
• 0:36
  from AI. It's not always perfect and
• 0:39
  sometimes it's given them some unrealistic expectations that
• 0:42
  that need to be toned down out here.
• 0:45
  And welcome to Cooper's code. We're here with our occasional
• 0:49
  AI episode with Marshall Cole and Sherrod Milnfar.
• 0:53
  Thanks for for joining. And Marshall, nice to have you
• 0:55
  back from from your trials. We're just going to dive right
• 0:58
  into stuff. We've got a lot of interesting
• 0:59
  things to cover today, everything from social media
• 1:02
  trials to prompting to pro pers using hallucinations to AI
• 1:08
  products in and the potential in ML cases there.
• 1:12
  So without further ado, I think Marshall, we'll, we'll kick it
• 1:17
  to you. Sounds like you were were given
• 1:20
  a talk at at the Santa Barbara Aboda recently and talking a bit
• 1:24
  about how AI is showing up as a potential maybe defendant in a
• 1:30
  case. Maybe you can unpack that a
• 1:31
  little bit, yeah. Yeah, sure.
• 1:33
  I mean, on this podcast, we've talked a lot about lawyers and
• 1:36
  how we're using AI and how we're getting it wrong and people are
• 1:39
  making mistakes and and all the things related to that.
• 1:43
  And they're sort of shifted the focus on, you know, we as
• 1:46
  lawyers now on behalf of your clients need to be looking for
• 1:51
  cases that are arising out of AI and it's misuse in other
• 1:55
  industries. And obviously, you know, every
• 1:59
  other industry is, is using it, probably using it or should be
• 2:02
  using it. And just like in the legal
• 2:04
  community, they're getting it wrong and messing things up as
• 2:08
  well. And so I was looking at a
• 2:11
  particular article regarding this nasal surgical device that
• 2:16
  was being used. And they had implemented
• 2:19
  artificial intelligence several years ago.
• 2:22
  And prior to implementing it, they'd only had 7 malfunctions
• 2:27
  or sort of botched cases, which would cause, you know,
• 2:31
  significant damage to to the person.
• 2:33
  But then since implementing AI, the malfunctions rose to in and
• 2:40
  being in excess of 100. And you know, the injuries are
• 2:45
  skull punctures, cerebral spinal fluid leaks, carotid artery
• 2:50
  damage. And so when you look at that
• 2:53
  type of a case, then when you start sort of unpack that and
• 2:57
  looked at, you know who you pursue, right, You've got the
• 3:00
  physician. Who?
• 3:01
  May or may not have used it appropriately.
• 3:04
  You've got the the AI manufacturer, you've got the
• 3:08
  hospital, the systems in place there.
• 3:11
  And so, you know, prior to the advent of AI, you would have
• 3:16
  arguably looked at that as a traditional medical malpractice
• 3:19
  case. You know, a lawyer botched a
• 3:21
  surgery, caused injuries and you know, is damaged and, and
• 3:26
  damaged the the patient. Now with AI involved, if that's
• 3:30
  a third party manufacturer, you've got a potential product
• 3:33
  liability claim. And as we were discussing
• 3:36
  offline, you know, with micro caps in relation to that apply
• 3:39
  to Med Mal cases, if you add a product liability component to
• 3:44
  it, you you move it outside of that realm and can recover a lot
• 3:48
  more money for your client depending on what the
• 3:51
  circumstances are. And this was just a pretty, you
• 3:54
  know, basic example of a case study where this was going on.
• 3:59
  And sure, if you are in this field, you don't have to scratch
• 4:03
  very deep into the surface to find other areas of medicine
• 4:07
  where AI is being used. I know it's definitely being
• 4:11
  used it for radiologists looking at scans and you know, I'm sure
• 4:18
  it's only a matter of time before people are starting to
• 4:22
  get it wrong by over relying upon AI and getting misdiagnosis
• 4:26
  or misdiagnosed or however it might be.
• 4:29
  So it was a really engaging discussion and just kind of a
• 4:32
  refreshing way to look at and think about as lawyers AI, not
• 4:37
  just how we're using it and shouldn't shouldn't be using it,
• 4:40
  but how all the consumers out there are using it and potential
• 4:44
  cases that exist. And then even on the defense
• 4:46
  side, you need to be looking out for your clients and advising
• 4:50
  them, just like, you know, we've been doing in the past episodes
• 4:54
  for lawyers, you know, for appropriate uses or responsible
• 4:58
  and ethical usage of artificial intelligence.
• 5:01
  Sharon, have you come across this in, in and, and I know you
• 5:05
  do a wide variety. I, I, I always think of you as
• 5:07
  the elder abuse guy. Give me your verdicts.
• 5:10
  But, but I know you do a wide variety of things.
• 5:12
  Have you started coming across this in ways where like you're,
• 5:15
  you're questioning, say APMK and finding that, oh wait, there's,
• 5:21
  there's actually a potential other defendant behind these
• 5:24
  defendants with an AI tool or, or anything along those lines?
• 5:28
  You know, I haven't had anything specific come up in that
• 5:30
  context, but just kind of putting my defense lawyer hat on
• 5:34
  from prior experience. I mean, especially if you're a
• 5:37
  defense lawyer defending a Doctor Who has what's known as a
• 5:40
  hammer clause and says I don't want to settle this case, it
• 5:43
  wasn't my fault. A great way to kind of get your
• 5:47
  client out of hot water is to find out if there was something
• 5:50
  else that caused the injury. And if that's something else is
• 5:54
  AI, which doesn't have a Med mouth cap attached to it.
• 5:58
  That would be a significant piece of information that could
• 6:03
  help not only your client but could also help the person who
• 6:07
  was injured. And these are things that
• 6:09
  Marshall is talking about, in my opinion, is that you have to
• 6:12
  look outside the box and kind of take the blinders off and always
• 6:17
  look for new things that can help our clients achieve their
• 6:20
  goals in litigation. It certainly feels like in
• 6:24
  taking depositions, Medmal, any other area good, some good
• 6:28
  open-ended questions about whether there was any.
• 6:31
  Was there any sort of AI assist, any tools, anything built into
• 6:35
  Epic? That's the underlying the
• 6:38
  majority of of hospitals and doctors use an Epic system where
• 6:42
  where there are some tools that reside within there in terms of
• 6:44
  their their medical charts. I think depo questions are a
• 6:47
  definite addition to the tool belt of a Med now lawyer anyone
• 6:53
  or it really anything. I mean, if you're deposing APMK
• 6:56
  of a finance company, I mean, you know, if they're, if they've
• 6:59
  implemented it, how it's how the algorithms are working, what are
• 7:02
  they doing? What are the recommendations
• 7:04
  based upon, you know, what's the human component and where does
• 7:07
  that intersect with AI? And inevitably, right now, at
• 7:11
  that cross point, something's potentially going wrong.
• 7:14
  Well, and I think we all have had experiences if you've
• 7:17
  handled any kind of trucking cases, I used to do tons and
• 7:20
  tons of trucking cases and downloading the, you know,
• 7:23
  electronic control module on those trucks.
• 7:26
  Now they have cameras that are facing the driver where there
• 7:30
  may have AI monitoring to see if this driver is getting and
• 7:33
  drowsy. And all of those things are
• 7:36
  things that we have to ask about and see.
• 7:39
  You know what's the worst that can happen?
• 7:40
  They didn't use it, but if they did use it, that could be
• 7:43
  significant information that could be useful in prosecuting
• 7:47
  your case in a trucking case or even any kind of a ride share
• 7:52
  type case that might come up. Anything further on this topic
• 7:56
  before I, I, I, I feel Marshall with you being in Los Angeles,
• 8:00
  I'm, I'm anxious to get to social media trials and, and
• 8:02
  Mark Lanier and Mark Zuckerberg and things along those lines.
• 8:06
  No, it was just, it was something that came up and I
• 8:09
  thought it was just an interesting thing for lawyers to
• 8:12
  start focusing on. And I'm sure many have.
• 8:14
  But you know, if handling your cases, you need to start adding
• 8:18
  questions in discovery, what you know regarding AI depos, cross
• 8:23
  examining witnesses, expert witnesses, the standard of care,
• 8:27
  right? It does it or does it not
• 8:29
  require the use of AII? Think it's very field dependent.
• 8:32
  But if you're not asking about it, I think you're just there's
• 8:35
  a big gap of information you're leaving behind.
• 8:37
  You don't know, you're not benefiting your client.
• 8:40
  So that was all I thought was it would be beneficial to kind of
• 8:42
  bring that up and just another way for us to think about using
• 8:46
  it. So the takeaways on on the the
• 8:48
  deposition checklist, make sure that there's a slot for AI now
• 8:51
  to see if there's any potential big broad open-ended questions
• 8:55
  to see if there's something there.
• 8:56
  Yeah. And then I and I think you need
• 8:58
  to in the particular field, as you narrow down those questions,
• 9:03
  you're going to need to become more educated to know the
• 9:05
  appropriate questions, right. I mean asking do you use AI?
• 9:09
  I mean, the person you're deposing may not even know.
• 9:12
  You got to define it. They've got to know what it is.
• 9:15
  That general term artificial intelligence is so broad that I
• 9:20
  don't think you'll get very effective and useful answers
• 9:25
  without knowing exactly what specific AI tools are being
• 9:30
  used, how they're being used. And I think the only way you
• 9:33
  learn about that is by learning more and more about the
• 9:37
  different artificial intelligence tools that exist in
• 9:41
  those various fields. Not to go too far off field, but
• 9:44
  that goes into how you take that positions and the funnel
• 9:47
  technique that's taught in the National Institute for Trial
• 9:50
  Advocacy programs, right, is if you ask what types of systems or
• 9:54
  what are the names of the systems that you use, you can
• 9:57
  almost get real time information on them because of the AI
• 10:01
  systems that are out there available to allow you to do
• 10:03
  research as you're asking the question.
• 10:05
  So it's kind of a two way St. in the sense that we want to know
• 10:10
  what they're doing, but also what tools are we using to get
• 10:13
  educated on those topics. And to your point, I think both
• 10:18
  your points that the physician or healthcare provider may not
• 10:25
  even be aware that there is some back end tool.
• 10:28
  It's you know, when the radiologist says, well, the you
• 10:31
  know, the computer told me it was likely non cancerous.
• 10:34
  You know, it, it's listening. You know what?
• 10:35
  Oh, what do you mean the computer told you?
• 10:37
  That's where the dig is. Yeah.
• 10:39
  Or if they even say take that answer, if they just say I
• 10:43
  didn't see it, right, Is it really them who didn't see it?
• 10:46
  And you have to ask that follow up question.
• 10:49
  I think most lawyers, rightfully so, would accept that answer,
• 10:52
  right? I didn't see it, I missed it,
• 10:54
  whatever it might be. But the eye is a big difference.
• 10:57
  Now there's there's a lot more there that you need to ask other
• 11:01
  than OK, accept that answer, move on, Doctor messed up,
• 11:04
  whatever. As we transition into the social
• 11:07
  media trial, I have to ask Sharad, are those glasses that
• 11:12
  you're wearing meta glasses? Not a chance.
• 11:16
  I, I bring that up only because from media management side, I
• 11:21
  have no idea how the trial is going to turn out.
• 11:23
  I will say Zuckerberg showing up being phalanxed by Meta glass
• 11:30
  wearing security seemed like a bad look going into this thing
• 11:37
  that that's just my my my kind of catty comment there.
• 11:40
  I I don't know if either of you have strong feelings about that.
• 11:43
  I saw something that's indicated that the judge noticed people in
• 11:47
  the audience wearing meta glasses and basically gave a
• 11:51
  very strong warning that any recording that was done in the
• 11:55
  courtroom must be deleted because they don't want the face
• 11:58
  of the jurors out there. And then one of the lead
• 12:02
  attorneys in the case also gave an interview in the first floor
• 12:06
  of the courthouse on Zoom, and that apparently violated the
• 12:10
  court's orders on no photography inside the courthouse.
• 12:14
  So there's, you know, and that's kind of goes to what we all talk
• 12:18
  about in terms of social media as well.
• 12:20
  But this trial has a number of different layers that, from just
• 12:24
  a human interest standpoint, I think we should all be watching.
• 12:28
  Definitely. And and I tell people I don't
• 12:30
  follow sports, I follow lawyers. And so when you look at somebody
• 12:33
  like Mark Lanier, who did the Vioxx first Vioxx trial and, and
• 12:36
  and just to kind of has a record of being one of the go to guys
• 12:40
  for the bellwether cases. I have to imagine that it it's
• 12:46
  quite an experience to watch him take a witness like Mark
• 12:51
  Zuckerberg or or the the head of Instagram.
• 12:54
  It did not sound like, at least from the from the digest version
• 12:58
  that we're getting from from outside the courthouse, that
• 13:00
  that those went particularly well for the defense.
• 13:04
  Yeah, I mean, I think the press was that Zuckerberg was very
• 13:07
  combative and indicating that he is being, you know, taken out of
• 13:10
  context and it's not what people think it is.
• 13:15
  But when your own internal documents, you have your, you
• 13:18
  know, your staff saying, you know, Instagram is a drug, and
• 13:24
  then the response says, yeah, and we're the pushers.
• 13:27
  That's a hard thing to defend regardless of who you are and
• 13:31
  how smart you are. And I don't know how well Mark
• 13:34
  Zuckerberg did. He seems like he's worked on his
• 13:36
  personality over the years. But having seen him testify
• 13:40
  before Congress it I would love to see the video of how that
• 13:44
  exchange goes with Mark Lanier, who is a very polished and
• 13:49
  incredibly prepared lawyer. One other piece that that I
• 13:53
  thought was interesting in terms of the examination and, and kind
• 13:56
  of a, a wise move to the extent that Zuckerberg was more
• 14:02
  polished and more capable than he has been in the past using
• 14:06
  the, you know, hundreds of hours of media training that he has
• 14:09
  been subjected to as as part of the cross.
• 14:13
  I I thought that was an interesting, interesting
• 14:14
  approach. I thought so too, and I think he
• 14:16
  acknowledged Like this isn't my strong suit, right?
• 14:18
  Like, I'm not that good at this. So, you know, take what I say
• 14:24
  with a grain of salt in some ways.
• 14:25
  And I think, you know, I understand why he said it.
• 14:29
  And I think that was good prep for him to say something like
• 14:33
  that. But like you said, Chirada, I
• 14:34
  think it's a tough position to defend.
• 14:38
  Wait. You know, he's on the one hand
• 14:39
  trying to do a good thing, offer service to millions and millions
• 14:42
  of people, and on the other hand, bad things are happening
• 14:44
  because of it. And he's at an intersection of
• 14:48
  profits versus sort of ethical issues.
• 14:53
  Well, I think that name is section is way in the rear view
• 14:55
  mirror. I mean they going back to
• 15:01
  Marshall's point in terms of looking at different ways to
• 15:03
  bring cases. This is an excellent example,
• 15:06
  right, because the Section 230 typically would bar any kind of
• 15:11
  liability for these companies. But yet because it's couched as
• 15:16
  a product liability case, that's been a very effective way of
• 15:21
  getting around that that particular statute.
• 15:23
  So again, it's, you know, Mark Lanier and I believe Rahul
• 15:28
  Ravaputi's involved in that case as well, are, you know, being
• 15:32
  very creative in how they're doing their job and able to get
• 15:37
  it to trial, which is not an easy task.
• 15:40
  As a parent who is, you know, I've got a 12 and 13 year old
• 15:45
  and, and who is seeing how addictive these tools are, I am
• 15:53
  cautiously optimistic that this kind of approach, if we can't
• 15:58
  rein in through Congress, perhaps we can rein in through
• 16:00
  dollars. Is is is is is part of my
• 16:02
  sentiment. I think it's the most effective
• 16:05
  way. I mean, I, I think the most
• 16:07
  enjoyable videos I saw were of Zuckerberg getting served with
• 16:11
  various other lawsuits as he was walking in and out of the
• 16:14
  courthouse. I mean, you know, these people
• 16:16
  aren't very accessible except in moments like this where they are
• 16:21
  rarely, you know, in a public place exposed.
• 16:24
  Who knows what they're being served with.
• 16:26
  But I thought that was just a little bit of a funny anecdote
• 16:30
  to take away from the whole. Thing that's going on, either of
• 16:34
  you have any predictions just based on what you've read in
• 16:36
  terms of the case? I think the cases like this are
• 16:39
  tough. I, I, I get it.
• 16:42
  I see the avenue to prevailing. I think it's, it is addictive.
• 16:47
  I think, you know, the age verification process, on the
• 16:50
  other hand, is hard for them to get right.
• 16:53
  But I just think when you look at the particular elements of
• 16:58
  what the plaintiffs need to prove, I think it's, it's still
• 17:01
  tough, you know, But so I, I, I, I see both sides of it, you
• 17:06
  know, from the, the moral side, the moral victory, why the
• 17:10
  plaintiffs should prevail here for the, all the right reasons.
• 17:13
  But then I think there's just the legal hurdles that they have
• 17:15
  to overcome to actually meet the burden of proof.
• 17:19
  Yeah. But that's why we have good
• 17:21
  lawyers on both sides and that's why these things are tried.
• 17:24
  I mean, if it was a slam dunk for one side or the other, you
• 17:28
  never go to trial. Yeah.
• 17:32
  I always say the juries, the juries get it right, you know,
• 17:34
  whatever that they get it right, so they'll get it right.
• 17:37
  That will be the right decision, whatever it is.
• 17:39
  I'm excited to see what the outcome is.
• 17:41
  It, it, it, it has been interesting to follow this case
• 17:44
  along the lines of cases. Flipping it back to, to, to you,
• 17:47
  Marshall, a couple months ago, I, I, I think November at, at
• 17:52
  this point, the, you had a pro pro who had done some motion
• 17:57
  practice with hallucinated cases.
• 18:00
  And I know that that was pending back then.
• 18:02
  And one of the things we said is we're going to return to it.
• 18:05
  And so I, I had a little footnote to make sure that we,
• 18:07
  we came back to that to see how that got handled by the judge
• 18:11
  and, and, and use that kind of as a launching pad in terms of,
• 18:14
  of AI experiences with, with people who are going the path
• 18:19
  without wares. So.
• 18:21
  Most of my practice is defending lawyers.
• 18:23
  And so for whatever reason, often times I deal with propers
• 18:28
  as plaintiffs who sue their, their former counsel.
• 18:31
  And, you know, at any given time, I probably in the office,
• 18:37
  we probably have, you know, 5 to 7 propers we're doing.
• 18:40
  And in this one particular instance, this proper, the case
• 18:44
  has been pending for a very long time and you know, obviously
• 18:49
  started using AI to write briefs and it was pretty obvious that
• 18:52
  that was happening. And it was only a matter of time
• 18:55
  before hallucinated cases that didn't exist.
• 18:58
  We're going to pop into those those motions which happened in
• 19:02
  one such instance, motion denied court stern, very stern ruling,
• 19:10
  but no real repercussions other than that.
• 19:14
  If it was a lawyer, it certainly would have been a lot worse.
• 19:17
  Or for myself, I had done it, it would have been a lot worse,
• 19:20
  right? It would have been another
• 19:20
  headline that we've seen dozens and dozens of at this point in
• 19:24
  time. But you know, separate from that
• 19:26
  and in another case that I've got with a pro per who started
• 19:30
  using it all across the board and in my e-mail communications,
• 19:34
  you could see overnight he figured out how to start using
• 19:37
  whatever it is, probably ChatGPT.
• 19:40
  And his emails sound very lawyerly.
• 19:43
  And I presume he's writing some a couple sentences and says make
• 19:46
  this sound like a lawyer. And sure enough, it does a
• 19:48
  pretty darn good job. And I've outlined in my response
• 19:52
  emails, you know, you don't need to use AI to draft your emails.
• 19:54
  It's perfectly fine. Just we can communicate.
• 19:57
  We were doing it before. There's no difference here.
• 20:00
  And it's, and it's so obvious. Just just stop.
• 20:03
  But he won't listen to me. That particular individual
• 20:06
  hasn't submitted anything with hallucinated cases or, you know,
• 20:10
  fabricated legal positions. It's all been accurate and.
• 20:15
  He he's done a much better job than he was previously doing and
• 20:19
  he's still going to run into a wall at some point.
• 20:22
  But I think, you know, for pro, pro litigants, they're going to
• 20:26
  have a major leg up if they figure out and know how to just
• 20:30
  check to make sure that the things they submit to the court
• 20:32
  aren't fake. They'll certainly take better
• 20:35
  depositions, which I know this gentleman will.
• 20:38
  When the time time comes, he's going to have an outline.
• 20:40
  He's going to ask questions that are going to be well formed
• 20:43
  that'll probably get around a lot of objections that otherwise
• 20:46
  would have been asserted in the past.
• 20:49
  And so, you know, I fully expect to see the pro pro world by and
• 20:54
  large improve with, you know, you know, 2 steps forward, one
• 20:57
  step backward kind of thing. But they're, they're going to be
• 21:01
  using it and are using it. And I see it on a regular basis
• 21:04
  now and it's mostly entertaining for me at this point.
• 21:07
  But it's definitely interesting to see, you know, they're
• 21:12
  there's the ones who file motions with accurate case law
• 21:17
  are far more convincing and compelling than they otherwise
• 21:21
  would have been. And you have to deal with them
• 21:22
  like that, like they really are a lawyer as opposed to kind of
• 21:28
  slapping around a little more when they didn't have any
• 21:31
  ability to, to put forth a good argue.
• 21:34
  So it's an interesting thing that's really just taking off.
• 21:38
  As AI has gotten more legal friendly, lawyer friendly, I'm
• 21:46
  finding that we're we're engaging with more educated
• 21:49
  consumers. What I mean by that is in the
• 21:52
  past in the personal injury world, often times somebody was
• 21:55
  coming to us with, with 0 experience and it and one of the
• 21:59
  nice things. Now it's, it's both nice and a
• 22:02
  problem is that they're much more educated.
• 22:05
  They they've done some digging on AI and have some sense they
• 22:09
  may have even drafted a demand package that they've sent off to
• 22:11
  the insurance company themselves before they've come to us.
• 22:14
  The the downside is the education that they've gotten is
• 22:18
  is is from AI. It's not always perfect and
• 22:21
  sometimes it's given them some unrealistic expectations that
• 22:24
  that need to be toned down. That's a launchpad for a
• 22:28
  question for, for, for both of you, perhaps Sharad, to start.
• 22:31
  Are you finding that you're getting more educated people or,
• 22:35
  or people who you can tell just in the conversation with them
• 22:37
  that they've they've done some digging before they're talking
• 22:40
  with you? Yeah, I think there's definitely
• 22:43
  an element of that that's appearing.
• 22:45
  It's just the part of what AI is doing it just from a social
• 22:50
  fabric perspective, is becoming almost a confidant to people,
• 22:54
  right? Which is what led to the suicide
• 22:57
  cases that we're reading about. And, you know, the toy case that
• 23:00
  I told you about, there's now AI toys for children where the kid
• 23:04
  can talk to the AI and literally have interactions as if this is
• 23:09
  their best friend. So I don't think that it's
• 23:12
  unusual for people who are in horrible situations, they've
• 23:16
  been in a, you know, horrific crash to think, all right, where
• 23:20
  do I start? Maybe AI can guide me in the
• 23:23
  right direction. So I don't blame them.
• 23:25
  It's just that there are potential pitfalls, especially
• 23:28
  if you're a criminal defendant, as in the case that we're going
• 23:31
  to talk about, and you decide to go to ChatGPT and put in a bunch
• 23:35
  of information that then the FBI gets through a search warrant
• 23:39
  that they execute on your house. Yeah, I've had clients of mine
• 23:43
  who are lawyers take work product of mine and I know
• 23:47
  they've put it through AI to edit it and revise it and send
• 23:49
  it back to me. And I've I've asked that because
• 23:52
  somebody comes out and, you know, when you see it enough and
• 23:54
  you use it enough, you can you still can really tell that the
• 23:57
  difference. Also the speed at which they've
• 23:59
  revised something and sent it back.
• 24:01
  And I asked them, you know, just that what you did and they all
• 24:03
  say yes. And I'm like, well, let let me
• 24:05
  tell you why. You know, we don't want to use
• 24:09
  that version completely. Now, certainly some suggestions
• 24:11
  come in and they're great and and they get implemented.
• 24:13
  But I I find that to be interesting that that's
• 24:17
  happening to me more and more often.
• 24:18
  The giveaway and and I I I mentioned this to Sherrod in an
• 24:22
  article recently. The most telltale sign to me
• 24:26
  that when somebody's handing me something in terms of draft form
• 24:30
  are and and. I won't remember the asking
• 24:32
  numbers for them, but the completely vertical quote and
• 24:37
  apostrophes the from the simple ASCII keyboard as opposed to a
• 24:41
  curved one. Those to me are dead giveaways
• 24:43
  that someone has pulled it out of an an AI editor because those
• 24:49
  don't show up in Word in the and and Google Docs in the same way.
• 24:54
  Maybe that's part of the prop now that you have to put in.
• 24:57
  Yeah, use word appropriate quotes please to avoid the
• 25:03
  scrutiny of Miles. There are a bunch of a bunch of,
• 25:09
  you know, college students taking notes now.
• 25:13
  So Sherrod, you had an interesting case on AI waiver
• 25:17
  and and and and search and, and so and you alluded to it.
• 25:21
  So let's let's let's go there now, if you don't mind.
• 25:23
  So there was a case in federal court in the Southern District
• 25:27
  of New York where the gentleman was indicted.
• 25:30
  And he then decided before going to his lawyers to put a bunch of
• 25:36
  information in an AI system, I think it was ChatGPT.
• 25:40
  And the FBI ended up executing a search warrant on his house.
• 25:45
  And they went in, took his computer and his phone and a
• 25:48
  bunch of other things that he had in his possession.
• 25:51
  And lo and behold, when they boot the computer up and start
• 25:54
  looking through his chat history on on the laptop, and I'm
• 25:58
  presuming he had the native native app on his computer or
• 26:04
  was logged in. So they were able to just
• 26:06
  basically go in and were able to obtain all this information he
• 26:11
  had on the case from those chats.
• 26:15
  And the court basically ruled that it was not protected by
• 26:18
  attorney-client privilege and it was not attorney work product
• 26:21
  because he wasn't consulting an attorney and he was just
• 26:24
  basically sending it out to a third party and not really
• 26:29
  exercising anything that was remotely A safeguard and
• 26:32
  protecting attorney-client privilege.
• 26:35
  So it didn't turn out too well for him in that context.
• 26:39
  So that's a pitfall, especially if you're a criminal defendant,
• 26:43
  that you have to be careful. About that's like a huge, you
• 26:47
  know, going back to sort of discovery and what we do as
• 26:49
  lawyers. I think if you have an adversary
• 26:53
  and whatever that you know Sake Tay say as an individual, right,
• 26:58
  that individual's communications with their lawyer are clearly
• 27:01
  privileged. You can't touch that.
• 27:03
  But if that individual takes those communications that it had
• 27:08
  with its lawyer, puts them into an AI platform to get a second
• 27:12
  opinion, OK, it has I now, I believe waived that privilege as
• 27:17
  far as those communications go. And I think as lawyers, you can
• 27:22
  propound discovery in relation to that individual and anything
• 27:28
  they've done with respect to 3rd party AI platforms potentially
• 27:34
  because you know, if they're going to take things like that.
• 27:37
  And I, I think it's a, a potential waiver problem because
• 27:40
  ChatGPT right on the website says anything you put onto yours
• 27:43
  not confidential, etcetera. So when I read that, that
• 27:47
  article we were talking about, that's that's the first thing I
• 27:49
  started thinking about. I think the way that the
• 27:56
  evidence code and discovery is written, Marshall, I believe you
• 28:01
  are correct. My, my, my upspeak or, or or
• 28:05
  thinking process conversation on it is it seems I'm, I'm trying
• 28:10
  to parse it a different way. You know, Socratic method back
• 28:12
  to law school. If somebody takes something and
• 28:15
  dumps it into Microsoft Word and re edits it, that's not
• 28:21
  discoverable. Even if you're using the 365
• 28:25
  plan where it's, it's going back and forth with the cloud.
• 28:27
  I mean, arguably, yeah, it it shouldn't be discoverable.
• 28:30
  So at what point does it become in, in terms of using a third
• 28:34
  party vendor, where's the line and and and in certain respects,
• 28:38
  where do we want the line to be? Yeah, I, I understand the
• 28:41
  distinction between word and, and, you know, and copilot or
• 28:45
  whatever else might be somewhat AI based versus Chachi BT, you
• 28:52
  know, open source platform, which specifically says it's not
• 28:56
  confidential. And then also, I think it
• 28:58
  depends on what you're using it for, right?
• 29:00
  If you're just using it to edit a document of this goes to where
• 29:03
  we want the line to be. You know, I can see that falling
• 29:06
  on the side of that's clearly, you know, not going to be
• 29:09
  discoverable or whatever the rule it might be.
• 29:12
  But if you start asking it for legal advice and 2nd opinions
• 29:16
  and things like that, I I think it starts to open things up a
• 29:20
  lot more. And now you got me into the
• 29:24
  world of whether a say legal focused update, because I know
• 29:29
  Anthropic has had a legal focused update recently.
• 29:33
  If you have a legal focused update and a subscription plan,
• 29:37
  do does it reach a point where it's it's attorney-client
• 29:42
  communication and at what point is it not a machine?
• 29:45
  Anymore, well, I think in order for there to be an
• 29:46
  attorney-client privilege, you got to have individuals, you got
• 29:49
  to have people, right? Taking it sort of back a couple
• 29:54
  steps and, and things that you need to ask for, You know, we're
• 29:58
  used to asking for all communications between one
• 30:01
  person and another person, right?
• 30:03
  Or one person in this particular business.
• 30:05
  I think now you need to ask for communications between a
• 30:08
  particular person and a particular AI platform or all of
• 30:12
  that, because I think employees before they submit memos now and
• 30:17
  things like that, they're they're asking AI, did I get
• 30:21
  this right? You know what I mean?
• 30:23
  Like, like really probably, I don't want to say dumb, but kind
• 30:27
  of dumb, potentially liability based questions that they aren't
• 30:33
  going to send to that person, their Co worker, but now they're
• 30:37
  sending it to AI to get confirmation that they're right
• 30:40
  or they're wrong or is it illegal or is it not illegal?
• 30:43
  And then they send the e-mail. And so I think, you know, again,
• 30:46
  going back to this creativity, you should be asking for those
• 30:49
  types of communications because I don't see how they can assert
• 30:52
  the privilege there that wouldn't be there.
• 30:54
  That's just more, you know, a whole nother avenue of of
• 30:58
  potential treasure trove of dockets.
• 31:02
  I'm just looking forward to deposing the Tesla, a robot
• 31:06
  that's going to be in everyone's homes in a few years, so.
• 31:09
  They probably have recorded a lot of things that yeah, yeah.
• 31:15
  Yeah. But I mean, it's this is the
• 31:16
  type of thing where the law I think is constantly chasing
• 31:20
  reality. And it there's going to be many
• 31:23
  more of these types of opinions that we're going to be talking
• 31:27
  about going forward because people will be very creative in
• 31:31
  how they use these systems and how, you know, others go after
• 31:35
  the information that may be there or another.
• 31:38
  Platform. I have to say after having read
• 31:39
  a little bit about Bernie Sanders talk down on the
• 31:43
  peninsula recently, I'm really not worried.
• 31:46
  Within 12 to 18 months all white collar jobs will be gone and and
• 31:51
  we'll just all be on universal income and that includes us.
• 31:54
  So I've actually already been picking out my cabana.
• 31:56
  I don't know if you 2 have been doing that.
• 31:59
  I haven't gone that far yet. We're waiting for your
• 32:03
  recommendations. If you can put that in the
• 32:05
  ChatGPT and send us a memo on it, that'd be great.
• 32:09
  It will be discoverable. Any other pieces, AI topics,
• 32:15
  tech, other things that that are exciting you or or or feel we
• 32:21
  should be paying attention to in this episode?
• 32:23
  I think just, you know, we focused a little more on
• 32:26
  discovery and, and sort of technical legal aspects.
• 32:29
  I, I think, and you mentioned the rules of evidence.
• 32:32
  I think, you know, watching out for amendments to the California
• 32:37
  and Federal Rules of evidence, there's already been proposals
• 32:40
  on the federal side with respect to how to treat AI based
• 32:45
  evidence. Hearsay rules are going to be
• 32:49
  redefined as AI evidence starts to get used.
• 32:52
  Authenticity, how you authenticate the various
• 32:56
  documents or videos or whatever you might be trying to use is
• 33:00
  going to become bit how for difficult with AI.
• 33:04
  But hearsay is probably going to be easier with AI because
• 33:06
  there's no declarant. So I think it's just important
• 33:09
  to think about those issues. If you have trials and you're
• 33:12
  using AI, if you, you know, accident reconstruction used to
• 33:17
  be is is still a huge thing right on your guys side and on
• 33:20
  both sides are reconstructing. They pay a vendor to do that who
• 33:24
  can come in and talk about it and authenticate it.
• 33:27
  Now, arguably that field could be one who would be maybe
• 33:31
  eliminated because you can have that accident reconstructed by
• 33:36
  artificial intelligence, but then you got to get it into
• 33:39
  evidence and who's your witness to authenticate it, You know, So
• 33:43
  got to be thinking about those things.
• 33:45
  And there's a lot of articles that are coming out very well
• 33:49
  written on the various topics that are relating to the federal
• 33:53
  and California rules. But finally, it's just important
• 33:55
  for us as lawyers to think about that as well as we start trying
• 33:58
  cases that involve. AI good boy.
• 34:02
  I only make good points so. I don't know I I've I've read a
• 34:09
  few AI written emails from your rapper who who disagrees with.
• 34:15
  You he has, he has a lot of not kind of things to say about you
• 34:18
  so. Well, anything else Sherrod
• 34:25
  Marshall, do you think we should cover in this episode?
• 34:28
  Well, just the one issue we talked about earlier, which was
• 34:31
  this toy that says stuffed animal called the company is
• 34:35
  called Bondu BONDU. Apparently they exposed 50,000
• 34:40
  plus chats to anybody who had an e-mail account, a Gmail account.
• 34:45
  So as a parent, be careful in terms of what you're handing
• 34:48
  your kids, especially if they're, you know, toddlers and
• 34:52
  what information those kids provide the AI in in interacting
• 34:58
  with it and how that can result in reliance on this stuffed
• 35:02
  animal. As if it's a human being which
• 35:05
  has a lot of layers to it that we would have to unpack.
• 35:08
  His parents buying a kid or grandkid a toy.
• 35:13
  Yeah, I, I, I try to avoid the stuffed animal AI enabled
• 35:16
  stuffed animals for my kids. I just give them loaded guns in
• 35:18
  the stead. Mine just get Lib Boo boos.
• 35:20
  I don't know what's worse. Lib Boo boos.
• 35:22
  Yeah, Yeah. You do direct deposit for that.
• 35:25
  Yeah, don't get me started. Thankfully, they're not.
• 35:28
  Connected to the Internet though, well, I, I, I think with
• 35:33
  that we're work we're coming to a close.
• 35:35
  I, I really appreciate our conversations.
• 35:38
  I always feel more educated and definitely entertained after,
• 35:42
  after one of our episodes. So thanks Marshall.
• 35:44
  Thanks, Sharad. He's, as always, glad to be
• 35:46
  back. And thank you all for listening.
• 35:49
  If you have any feedback on any topics you'd like us to cover on
• 35:52
  on the AI Show, please feel free to reach out at podcast at
• 35:56
  Coopers dot Law. And to all you doing justice out
• 35:59
  there, happy hunting.