Latest / CLEs You Actually Want to Hear / AI in Alternative Dispute Resolution and the Practice of Law
Transcript
- 0:02Hello everyone and welcome to today's program,
- 0:06AI and Alternative Dispute Resolution and the
- 0:09Practice of Law. I'm Alla Policastro, Director
- 0:12of Education and Professional Development at
- 0:14Beverly Hills Bar Association. And today I'm
- 0:17delighted to introduce our speakers, Elizabeth
- 0:19Rothman and Ryan Abbott. Elizabeth is an attorney,
- 0:24author, and thought leader at the nexus of law
- 0:27and advanced technology. Her legal career is
- 0:30marked by a deep engagement in intellectual property,
- 0:33contracts, healthcare, and the evolving technological
- 0:36landscape. She, along with our second presenter,
- 0:39Ryan Abbott, are co -authors of the Elgar Concise
- 0:42Encyclopedia of Artificial Intelligence and the
- 0:44Law, published in 2025. And she publishes regularly
- 0:49on intellectual property, generative AI, and
- 0:51legal and policy issues surrounding technical
- 0:55innovations. Elizabeth is an adjunct professor
- 0:58at the University of Southern California Gould
- 1:00School of Law and the University of Illinois
- 1:02College of Law. And she's on the faculty of PatentX,
- 1:06a certificate course on patent law and global
- 1:09public health offered jointly by Harvard Law
- 1:11School and the World Intellectual Property Organization.
- 1:15The full extent of her resume could keep us here
- 1:17for the entire hour. So I invite you to review
- 1:20her resume and her qualifications in the materials
- 1:24packet that you've received before this program.
- 1:27Dr. Ryan Abbott is a licensed physician, attorney,
- 1:31and mediator and arbitrator with JAMS. He has
- 1:34extensive experience in intellectual property,
- 1:37healthcare, with expertise in areas including
- 1:39biologics, pharmaceuticals, medical devices,
- 1:43life sciences, computer technology, and distributed
- 1:46ledger technology. Dr. Abbott has served as a
- 1:49partner at a litigation firm, general counsel
- 1:52of a biotechnology company, and of counsel at
- 1:55a firm specializing in transactional matters
- 1:57and intellectual property litigation for life
- 2:00science companies. He has written widely on issues
- 2:03associated with law and technology, health law,
- 2:06and intellectual property in leading legal, medical,
- 2:09and scientific publications. He's also professor
- 2:12of law and health sciences at the University
- 2:14of Surrey, as well as adjunct associate professor
- 2:16of medicine at the University of California,
- 2:19Los Angeles. And now I'm going to turn it over
- 2:21to them so they can get started. I'm just going
- 2:25to share my screen real quick. Thank you so much
- 2:27for that introduction. And I'm very happy to
- 2:30be here today with the Beverly Hills Bar Association.
- 2:34I'm going to start off by talking a little bit
- 2:36about AI and the legal profession today, an ever
- 2:39evolving topic that kind of is just constantly
- 2:42moving and changing and a very exciting area
- 2:45to be in. As was just mentioned, Ryan and I are
- 2:51co -editors of the Elder Concise Encyclopedia
- 2:53of Artificial Intelligence and the Law. So I
- 2:55encourage you to check that out. It published
- 2:57at the end of last year. It is a concise reference
- 3:00work, so the first of its kind in recent years
- 3:04to bring together this intersection of AI and
- 3:07the law in this comprehensive way spanning across
- 3:11100, I think, 107 entries with over 100 experts
- 3:14authoring these entries from all over the world.
- 3:18And so it was a very, very exciting project to
- 3:21come to completion and to see in print and just
- 3:25came out at the end of last year. So the insights
- 3:28learned from the encyclopedia and that this area
- 3:33of law is very, very rapidly evolving. And the
- 3:38implications of that for the legal world are
- 3:40constantly shifting. And understanding those
- 3:45implications requires deep collaboration between
- 3:47technologists and lawyers, especially where there
- 3:52are regulatory approaches varying significantly
- 3:54around the world and changing frequently. We
- 3:58really have to think thoughtfully and consider
- 4:03how to take these complex AI concepts and bring
- 4:07them into legal guidance that is functional for
- 4:09our profession. And that is an evolving and kind
- 4:14of ever difficult challenge that we're going
- 4:17to be facing going forward as a profession. So
- 4:21AI touches pretty much every area of law. I feel
- 4:24like I should have put a little asterisk in here
- 4:25because I'm sure there's something. But it really
- 4:28is at this point touching every area of the law
- 4:31that I can think of, and especially with governments
- 4:35and large entities around the world using AI.
- 4:39So frequently in the public adopting AI as well.
- 4:41Really, there are not many areas of our lives
- 4:44that are untouched by AI at this point. So even
- 4:51if you don't use AI, this still matters. And
- 4:55your clients are living in a world that is being
- 4:58shaped by AI, you are living in a world that
- 5:00is being shaped by AI, even if you're not using
- 5:02AI tools in your practice. AI is being used by
- 5:05courts, opposing counsel, government agencies,
- 5:08insurers, your clients, employers. Opposing counsel
- 5:13might use AI tools that affect discovery, evidence,
- 5:16jury selection, litigation strategies. Courts
- 5:20are across the country issuing using AI specific,
- 5:27issuing AI specific rules and things that you're
- 5:29going to need to be aware of in the practice
- 5:31kind of no matter what area you practice in and
- 5:35whether you use tools or not. The competence
- 5:39under model rule 1 .1 now includes understanding
- 5:42the benefits and risks of relevant technology.
- 5:45So it's really part of our professional duty
- 5:47to understand these technologies even if they're
- 5:49not being used directly in your practice. I have
- 5:54referenced a number of reports in these slides
- 5:57and referencing reports on AI use is really difficult
- 6:01because everything is evolving so quickly. These
- 6:04reports are from 2025 and some actually that
- 6:08just published this year, but even so the numbers
- 6:10are kind of. continually increasing generally,
- 6:14but the concerns are a moving target as well.
- 6:17So I've just referenced these studies here so
- 6:19you know where I got some of the numbers from.
- 6:22And I believe these slides are available to you.
- 6:24So if you'd like to take a closer look, please
- 6:27feel free to do so. So AI matters for law firms.
- 6:35The reason that it matters is that nearly 7 in
- 6:3810 professionals, according to these studies,
- 6:40now use generative AI tools for work in some
- 6:43capacity. That's more than double what it was
- 6:46last year in just a single year. In large firm,
- 6:51that figure can reach 87%. There's a lot of incentivization
- 6:55around using AI in large firms. And that is just
- 7:00certainly showing up in the numbers. The top
- 7:05uses of generative AI in legal practice are drafting
- 7:16correspondence. About 58 % of respondents said
- 7:20that they were drafting correspondence, legal
- 7:22research, brainstorming ideas, document summarization,
- 7:27and over 80 % of organizations generally using
- 7:32some kind of generative AI, at least weekly.
- 7:36So this generative AI use among firms has nearly
- 7:39doubled in the last year to 40 % of firms. Many
- 7:42firms are now formalizing AI policies and governance
- 7:47structures to manage those risks and ensure compliance.
- 7:50And I will go through some of the things that
- 7:51you might want to consider when you're looking
- 7:53at your own firm's AI policy or putting together
- 7:56one from the beginning. There are a number of
- 8:05AI tools that you've probably heard of if you're
- 8:07on this webinar. Harvey and Legora, some of the
- 8:11largest two dominant legal specific AI platforms
- 8:14being adopted by large law firms. Harvey was
- 8:17founded in 2022 and Legora in 2023. And they
- 8:23have a very large share of the market of large
- 8:26law firms. Thomson Reuters or Westlaw Co -Council
- 8:30is embedded into Westlaw. A lot of us are familiar
- 8:34with those platforms. The idea is that these
- 8:39models are grounded in the proprietary information
- 8:43of Thomson Reuters and therefore may have less
- 8:47hallucinations, although there have been some
- 8:49studies, and I'll talk a little bit more about
- 8:51it, where there are still found to be hallucinations
- 8:55in these models. So you still have to carefully
- 8:58review the output. Similarly with Lexis, Lexis
- 9:02plus Protege, their AI product and claims to
- 9:07have less hallucinations. And in a 2025 Stanford
- 9:11study, Showed that there were still hallucinations.
- 9:14But again, these are these numbers can move and
- 9:16change and there are not really there's not a
- 9:19lot of new data from these companies. So just
- 9:21being skeptical, you know, with your own use
- 9:23of it and making sure that you're double checking
- 9:25everything is important. And then there are the
- 9:28more generic tools of Chat GPT, Copilot, Gemini,
- 9:32Claude. Many lawyers, especially in smaller firms,
- 9:36use these general purpose AI tools. They have
- 9:39the highest hallucination risk and the fewest
- 9:41legal safeguards involved. So very, very important
- 9:45if you're using any of those to be very careful
- 9:48about client information and double checking
- 9:51everything that comes out of the models. There's
- 9:56something like Claude's legal plugin, which is
- 9:58in a kind of research preview mode. It is a file
- 10:01-based plugin, so it's not a full legal platform.
- 10:04There are a lot of different platforms that lawyers
- 10:06are turning to and using for legal work or for
- 10:12drafting or for either correspondence or more
- 10:15significant drafting. So all important things
- 10:17to be aware of. Some of the promises of AI are
- 10:24that it will streamline legal research and memo
- 10:27drafting, automated contract review and due diligence
- 10:29review, faster case management and deadline tracking.
- 10:33These are some of the more modest promises of
- 10:38AI. And then we have the perils on the other
- 10:42side. We have to look out for AI hallucinations,
- 10:45generating false citations, fabricated precedents,
- 10:49making up statutes. privacy risks that might
- 10:52threaten client confidentiality or attorney -client
- 10:55privilege. And maybe more significant for the
- 11:00long -term, over -reliance on these tools could
- 11:04possibly impact professional judgment and accountability
- 11:07going forward. So there are alarming rates of
- 11:17errors. AI models that are used for legal work.
- 11:21Again, these studies are, there is not a lot
- 11:25of new data. There's not a constant update of
- 11:27the hallucination rate for these, especially
- 11:30the larger platforms. So this 2025 study did
- 11:34find that Lexis and Westlaw AI Each hallucinated
- 11:38more than 17 % of the time. And the general purpose
- 11:42LLMs were hallucinating about 58 % of the time,
- 11:46which is very, very significant that there's
- 11:49errors or made up information in the output.
- 11:55Sanctions on attorneys are increasing as well.
- 11:57And there are over 1 ,000 cases now of documented
- 12:00hallucinations being submitted to courts. Hallucinations
- 12:09are not so easy to catch. Large language models
- 12:13are not databases that are retrieving facts like
- 12:16we're used to with traditional research platforms.
- 12:19They are statistical prediction engines. They're
- 12:21trained to generate the most probable next word
- 12:24given what came before. And so the output kind
- 12:28of looks right by design. And the case names
- 12:31will follow real naming conventions. The fake
- 12:34citations will use real court abbreviations.
- 12:37It all sounds very plausible. But at the end
- 12:42of the day, it is not accurate. And so these
- 12:45fabricated holdings can track logical structure.
- 12:48Sometimes it's very difficult to catch, even
- 12:51for seasoned attorneys. Also, errors might cluster
- 12:56in places that you don't expect. So research
- 12:59shows that hallucinations are not always random.
- 13:01They're more likely when the correct answer is
- 13:04obscure or jurisdiction specific. The question
- 13:07involves older or less digitized case law. Or
- 13:11the model is asked about a non -US system or
- 13:13information that it wasn't trained on as well.
- 13:18Or if the prompt is phrased very confidently
- 13:21or in a direction that might skew the answer
- 13:24coming out, then you might get a hallucination
- 13:27that is. moving in that direction of what the
- 13:30prompt was asking for if it is phrased very confidently.
- 13:35So these are all sort of things to consider if
- 13:37you're using AI or your clients are using AI
- 13:41or opposing counsel in this environment that
- 13:45we're in. And then there are the data security
- 13:50and confidentiality risks with putting sensitive
- 13:54client information into a third party platform.
- 13:59So the terms of service are a great place to
- 14:03go at first to figure out if an AI vendor is...
- 14:08retaining any of your information, if they're
- 14:10analyzing it, if they're sharing client data
- 14:13that is input into the system, because obviously
- 14:15that can have very significant ramifications
- 14:18for attorneys if that data is disclosed. Even
- 14:22if they have very tight data security procedures,
- 14:28there is always the possibility of some kind
- 14:29of breach. Vendor terms of service might actually
- 14:34even permit training on AI models on your clients'
- 14:38confidential documents or your confidential documents.
- 14:41So firms really have to vet the providers very,
- 14:45very well. And sometimes that means not not taking
- 14:48what they're saying at face value, but digging
- 14:51a little bit deeper into their compliance procedures
- 14:54and encryption and their guarantees around not
- 14:57retaining data. What I have found is that some
- 15:00vendors or representatives might overstate their
- 15:02security of AI tools. And sometimes that's out
- 15:05of a genuine misunderstanding for attorney obligations.
- 15:08No one's going to care as much as we do about
- 15:10the security of client data. And sometimes there's
- 15:14just a misunderstanding or just they're not thinking
- 15:17about it as deeply as they need to be. And sometimes
- 15:21just out of commercial interest, they might overstate
- 15:23their security and data procedures. There's a
- 15:30formal opinion by the ABA in 2024 about updating
- 15:35compliance rules that lawyers must understand
- 15:37AI's capabilities and limitations. I won't go
- 15:40through all of these, but they are in the packet
- 15:44that you've received and understanding generally
- 15:48the technology that is involved here and maintaining
- 15:52your current professional responsibility standards.
- 15:57with this new technology. The California bar
- 16:03also issued guidance on generative AI in 2023.
- 16:08Basically applies the existing California rules
- 16:12of professional conduct covering confidentiality,
- 16:14confidence, supervision, and candor directly
- 16:16to the use of AI. And then just this last couple
- 16:20of months, there is a SB574 was passed. that
- 16:28aim to convert existing bar guidance into enforceable
- 16:31statutes. So look out for that coming soon. There
- 16:38are other laws in California and federally that
- 16:43you might want to be aware of. SB 53, the first
- 16:48enacted AI safety statute requiring large AI
- 16:52developers to implement safety protocols. And
- 16:55then there is a tension between the federal government
- 16:58and states like California that are trying to
- 17:01regulate AI at the moment. And a lot of executive
- 17:06orders. Prioritizing the removal of barriers
- 17:10for AI development and so trying to open up the.
- 17:15remove any kind of regulation, which includes
- 17:18these state regulations coming from New York,
- 17:21Colorado, and California primarily. So you can
- 17:27take a look at some of these bills as well. Some
- 17:35interesting news about AI in the courtroom. There
- 17:39was a case in New York just a couple of weeks
- 17:41ago about AI -generated documents not being privileged
- 17:44because the client had initially drafted a document
- 17:47using a publicly available AI model. There are
- 17:50also a number of cases around sanctions for hallucinated
- 17:55case names showing up and the supervising attorneys
- 17:58being held responsible as well. AI agents are
- 18:08another area that has become the kind of new
- 18:13buzzword of the moment. Some concern here around
- 18:17giving too much access to AI agent platforms.
- 18:20An AI agent, unlike a chatbot that responds to
- 18:24prompts, an AI agent executes multiple tasks,
- 18:27usually autonomously. So they might send email.
- 18:29They might retrieve or update records. They might
- 18:33connect to external systems, file documents,
- 18:37make other conditional decisions. And often,
- 18:40that is with minimal oversight in between actions
- 18:43that are being taken. So the lack of monitoring
- 18:46or potential kill switches to turn off the agents
- 18:49is a problem. If you deploy an agent into a system,
- 18:52can you cleanly shut it off and take it out of
- 18:57the system, or is that not possible? If agents
- 19:00are given broad access to confidential information,
- 19:04then that agent autonomously transmits that data
- 19:07somewhere else. There could be issues there with
- 19:10unauthorized disclosure. And then taking a good
- 19:13look at your contracts with these vendors and
- 19:16others would be important if you're instituting
- 19:18any kind of AI agents into workflow, because
- 19:22most law firm contracts were written for more
- 19:25passive and predictable software and not taking
- 19:27into account some of these agent issues. As I
- 19:40mentioned, around 70%, maybe over that, of professionals
- 19:44are now using some kind of generative AI for
- 19:47work at some point, depending on which study
- 19:49you're looking at. That number has risen significantly
- 19:54over the last year, and I would expect that it
- 19:56will continue to rise with more tools being available
- 20:00and incentives to use them at a lot of larger
- 20:04firms. And around 46 % of firms currently, according
- 20:09to this most recent study, have some kind of
- 20:12governance policy around the use of AI. This
- 20:15number has also gone up significantly, but the
- 20:19governance of the technology is not moving as
- 20:21fast as the adoption of it is by individual attorneys.
- 20:29So if you're looking at your AI use policy or
- 20:32looking to create one from the start, some good
- 20:36things to consider are looking for approved use
- 20:39cases. So define what the permitted AI applications
- 20:43are. Is it for research, drafting, summarization?
- 20:47Are you using it for more high -risk things?
- 20:49And if so, have you vetted the platform as well
- 20:53enough to use it for those purposes? If you might
- 20:58choose to explicit any kind of use of certainly
- 21:01any kind of unsupervised client facing output,
- 21:05so things going directly to a client from an
- 21:08AI system would be, at this point, something
- 21:13you'd want to exclude. Data governance of classifying
- 21:17sensitive data, applying minimization principles,
- 21:19and setting clear retention and deletion rules
- 21:21for data that's processed by AI systems. requiring
- 21:25end -to -end encryption, having some rule -based
- 21:28access controls, and monitoring AI platforms
- 21:31across the firm, and then having a significant
- 21:36human review protocol in place where you're mandating
- 21:39attorney review of all AI output before use,
- 21:42and establishing some clear error correction
- 21:45and escalation procedures when hallucinations
- 21:47are detected. As I mentioned, the hallucinations
- 21:50are often very, very difficult to catch. Inevitably,
- 21:55if there's a lot of AI use in a law firm, some
- 21:57hallucination is going to get through. And having
- 22:00a procedure to detect and rectify these situations
- 22:05is important. So just some tips on developing
- 22:12an AI policy, assembling the right team. This
- 22:15is an interdisciplinary. It is not, it should
- 22:20not just be delegated to an IT team. It really
- 22:23has to have the attorneys involved thinking through
- 22:25all of the different angles of what could go
- 22:28wrong with the deployment of these systems. Auditing
- 22:34your current AI use can also be very, very helpful.
- 22:37So before drafting, find out what's actually
- 22:39happening in your law firm. What are attorneys
- 22:44using AI for right now? And then you can tailor
- 22:46your policy around that. And what are things
- 22:48that they might start to use it for to make sure
- 22:51that you actually have policies that are usable
- 22:53for your attorneys? And then calibrating the
- 22:58risk profile. So a solo practice policy will
- 23:01look very, very different from a big law firm's
- 23:03policy. And drafting it in plain language. We
- 23:05all know that if nobody can understand the policy
- 23:09or remember it, then it won't be followed. So
- 23:13making sure that you're using plain language
- 23:15for a policy like this and not too much technical
- 23:18jargon so that everybody knows what the rules
- 23:20are to follow. And then finally building in a
- 23:24review cycle. AI is moving faster than technology
- 23:30regulation is. So building in a mandatory cycle
- 23:33for the policy itself is an important thing to
- 23:38do. So as I mentioned, the hardest part of AI
- 23:42governance may not be just writing the policy,
- 23:43but getting the lawyers to follow it. And so
- 23:47modeling good use from leadership, having role
- 23:51-specific onboarding, again, really thinking
- 23:53through what kinds of tasks people are using
- 23:56AI for right now and what they might move into
- 23:59later, and then figuring out how to train for
- 24:02those purposes, integrating into existing workflows.
- 24:06And then this point of positive reinforcement.
- 24:10There's been a lot of mixed messages around the
- 24:13use of AI for the last couple of years. And if
- 24:15it is allowable, if it's too risky, a lot of
- 24:21lawyers are choosing right now that they want
- 24:23to move in the direction of using more AI. So
- 24:25having a kind of culture around the use where
- 24:28it's not necessarily a something that has to
- 24:32be hidden or something that is that they shouldn't
- 24:34be doing. But if you're going to incentivize
- 24:36it, then kind of creating some positive reinforcement
- 24:40as well, as opposed to just talking about cases
- 24:43where hallucinations have, you know, there's
- 24:45been sanctions for hallucinations being submitted
- 24:47to a court, but finding the positive parts of
- 24:50it as well. Implementation failures, having a
- 24:55policy without training. outdated approved tool
- 24:58list. So if your approved tool list is very old
- 25:00and everybody wants to use the newer tool that
- 25:03is the it tool of the moment, making sure that
- 25:07your tools are updated on there. Make sure you
- 25:10have an escalation path when there are errors
- 25:12found. And then keeping records can also be very,
- 25:17very important in figuring out where things have
- 25:19gone wrong if there is a problem. OK. And this
- 25:25is just about responding to incidents that happen.
- 25:32Client communication, considering if you're going
- 25:33to, you should actually disclose to clients,
- 25:37if you're using AI platforms, what you're using
- 25:39it for, maybe putting it in your engagement letter,
- 25:42defining a scope of AI use, oversight procedures,
- 25:45and even a right for clients to opt out of AI
- 25:49use. And finally, AI is here to stay. Strong
- 25:58governance of AI should not be a burden. It's
- 26:01a differentiator. Making sure that you have these
- 26:03policies in place is very important to build
- 26:06trust with clients. And for the near future,
- 26:09AI is more likely to augment lawyers than replace
- 26:12them, in my opinion at least, regardless of what
- 26:14you have heard. And so I think having policies
- 26:18and really governing the AI that you're working
- 26:21with is the most important thing at this point.
- 26:25So our role as professionals, staying informed,
- 26:28staying on top of recent guidance, advocating
- 26:30for governance and clear policies at your firms,
- 26:34and then leading with client first decisions
- 26:36about using AI. If it doesn't protect or serve
- 26:40your clients to use AI for a certain purpose,
- 26:42then consider if it's something that you should
- 26:44be doing. And so AI is, if it serves your clients,
- 26:51use it. If it doesn't, don't. But be aware of
- 26:53all of these things and make sure that you know
- 26:57that other people are using it as well if you
- 26:59choose to go the route of not using it. If you
- 27:02do use it, oversight is not negotiable. Meaningful
- 27:07human review is not negotiable. You have ethical
- 27:09obligations that will fall on you. And make sure
- 27:13that you understand the technologies enough to
- 27:17know what's happening in the market. And there's
- 27:20some sources at the end. And now I'm going to
- 27:23hand it over to Ryan. Thank you. Well, thank
- 27:28you. It's great to be here with all of you. And
- 27:31I recognize many of you in the participant list
- 27:34and a not unsurprising number of AI bots that
- 27:38are here taking notes on behalf of the human
- 27:41overlords who are either not paying attention
- 27:43or just really want to re -listen to this without
- 27:46re -watching the video. The good news is I think
- 27:49for everyone who's using an AI bot, The Beverly
- 27:51Hills Bar Association has promised a free hard
- 27:54copy of this encyclopedia to everyone. So look
- 27:58out for that. It is outrageously expensive and
- 28:01already out of date, having been published in
- 28:04December of last year. But not to worry, we're
- 28:08working on a sequel. I'd like to shift gears
- 28:11and focus a bit more on AI and ADR as I am a
- 28:15neutral at Jam. So it's a topic that's near and
- 28:18dear to my heart. And I'd like to start with
- 28:20a couple of videos. There's all sorts of ways
- 28:24of measuring AI progress and leave it to the
- 28:27internet to decide that one of the interesting
- 28:29metrics is what happens when you prompted generative
- 28:32AI system with Will Smith eating spaghetti. This
- 28:35is what happened in around 2022, 2023. And as
- 28:39you can see, the results are somewhat less than
- 28:42realistic. Here is one from about a month ago.
- 28:53That's the good stuff. And that's a wrap. Progress.
- 29:00Progress. technologically, which is all well
- 29:03and good. But for lawyers, all sorts of interesting
- 29:07impacts this is going to have. Economic and financial
- 29:11ones. The writer's strike now of a couple of
- 29:13years ago, though who knows it may be coming
- 29:15back, was explicitly about the use of AI in the
- 29:19creative industry. It's a whole other long topic.
- 29:23You know, as Liz alluded to, people are saying
- 29:26AI is going to replace all lawyers. People have
- 29:28been saying that for a very long time, and I've
- 29:30still got a job. But we'll see what you think
- 29:32when we get to the end of this presentation.
- 29:35And then all sorts of substantive legal disputes
- 29:39that are likely to emerge around these things
- 29:41in obvious and some not so obvious ways. I'll
- 29:45run through some of that as a bit of groundwork
- 29:47for our AI and ADR discussion. You know, one
- 29:51of which is synthetic content. So if you can
- 29:53make a video of Will Smith that isn't perfect
- 29:56and you can tell it's AI if you're looking closely,
- 29:59but nothing like it was a couple of years ago.
- 30:02Not a new problem. Synthetic content's been around
- 30:06for a long time. After Abraham Lincoln's assassination,
- 30:10people wanted more presidential -looking videos
- 30:13of him, or pictures of him, and didn't think
- 30:17they had enough. So they took an engraving of
- 30:20John Calhoun, his biggest political adversary,
- 30:23removed his head, put Abraham Lincoln's head
- 30:26on it, changed the title of some of the text
- 30:28there, and voila, you've got a best -selling
- 30:31portrait. But 150 years ago, it took a long time
- 30:37and a lot of technical skill to do something
- 30:39like that. And now I can do it in 30 seconds.
- 30:42And it will at least fool my uncles and aunts
- 30:45on Facebook. So this can be used commercially
- 30:48if I want Will Smith to promote my encyclopedia,
- 30:51which I have it on good authority that he does.
- 30:54If I want to use Will Smith for political disinformation.
- 31:00The possibilities are endless and problematic.
- 31:04And there's all sorts of interesting IP issues
- 31:07associated with these as well, like can you own
- 31:11that video of Will Smith, leaving aside the name
- 31:14likeness and image issues. People have been using
- 31:18AI to make creative works for a very long time,
- 31:22since the 50s and 60s. And the U .S. Copyright
- 31:25Office has had a policy since 1973 that if you
- 31:28use generative AI, basically, you can't own its
- 31:32output. And this has been the subject of some
- 31:35interesting cases, one that just had a cert petition
- 31:37denied to the Supreme Court and one going on
- 31:40in this record of Colorado right now. But just
- 31:43to give you a sense of how the technology is
- 31:45keeping up, you know, I can go into Google Gemini
- 31:48right now and say, generate a creative image
- 31:50and it will make a cat cooking a souffle. So
- 31:54a few years ago there was a factual dispute about
- 31:57whether AI could make a creative work autonomously
- 32:00or at least kind of stepping into the shoes of
- 32:02a human creative. Now you see, well, we passed
- 32:06that benchmark a long time ago. So, you know,
- 32:09you can debate whether this is art or creative
- 32:12for all sorts of reasons. Copyright doesn't traditionally
- 32:15make that sort of distinction. The doodle I'm
- 32:18doing while I'm talking. were I making a doodle
- 32:21and, you know, a great work of art. Get 70 years
- 32:24of protection plus my life. That's just the way
- 32:27copyright law was set up. But all sorts of interesting
- 32:30issues here, including, as we're in Beverly Hills,
- 32:33or at least virtually, as music and movie studios
- 32:37and creatives increasingly use AI and complex
- 32:40ways of human -AI interaction, lots of challenging
- 32:44ownership issues associated with that. And not
- 32:48something that has escaped the attention of at
- 32:50least the creatives who are writing about it.
- 32:52or the policymakers passing new laws about it,
- 32:56which Liz referred to. Lawyers have gotten into
- 32:59the act. You've heard some guidance on this.
- 33:01You don't want to be sanctioned for using generative
- 33:03AI. I did ask generative AI here about lawyer
- 33:06sanctions, but lawyers cannot seem to help themselves
- 33:10from having Chad GTP write a brief and then signing
- 33:13that brief without reviewing it. And I get the
- 33:15temptation and allure of it, but that has never
- 33:17been a permitted process. Whether you're having
- 33:20a summer associate, Or your romantic partner
- 33:24who's not a lawyer or your dog write your brief
- 33:26for you You have to review every citation before
- 33:29it goes in But there is no shortage of lawyers
- 33:32who continue to do this having sat through this
- 33:35webinar. You're not going to be one of them People
- 33:38are developing all sorts of interesting new interactions
- 33:42with AI So even people having romantic relationships
- 33:46AI with AI was not a super new thing, but the
- 33:50prevalence of it is certainly changing dramatically.
- 33:54And one of the most common uses of generative
- 33:56AI is for emotional support now. So people are
- 34:02interacting with AI in more and different ways
- 34:05than they have historically. And this has led
- 34:07to some negative outcomes from people suing ChatGDP,
- 34:12for example, for giving them bad legal advice
- 34:14or bad medical advice or for inducing them to
- 34:20commit suicide or for all sorts of negative possible
- 34:23outcomes. And of course, probably the most litigious
- 34:28current AI specific issue is this issue of copyright
- 34:32on the training side of things, though lots of
- 34:35claims being alleged all over the world, mainly
- 34:38in the United States. We're now up to, I can't
- 34:41remember the most recent number, but more than
- 34:4360 lawsuits against generative AI companies.
- 34:45broadly speaking, by right holders, some of which
- 34:49have found fair use, some of which haven't found
- 34:51fair use, some of which have resulted in multi
- 34:53-billion dollar settlements. So that's not going
- 34:56anywhere. It is crazy to me that as far away
- 35:02as one could have seen this coming, there aren't
- 35:04clearer rules and guidance on fair use and copyright,
- 35:08such that apparently you have at least 60 cases
- 35:10going on. But there we have it. And so the exciting
- 35:14news from all of this is more disputes, so more
- 35:19work for lawyers, which I assume most of you
- 35:21are, and more work for alternative dispute resolution
- 35:24professionals. And thus, we get to kind of the
- 35:29use of AI in dispute resolution. And there are
- 35:32a couple of levels on which this is occurring.
- 35:36So one is AI in the substance of a dispute rather
- 35:39than AI in the process of dispute resolution.
- 35:43And I together with another neutral at Jams,
- 35:45Daniel Gehry, co -created the Jams AI dispute
- 35:48resolution clause and rules. And this is in recognition
- 35:52of the fact that there are several specialized
- 35:55practice areas in ADR. And there can be benefits
- 35:58to having rules tailor made to those. JAMS has
- 36:01lots of different rules for lots of different
- 36:03things, employment rules, commercial rules, complex
- 36:07rules, distributed ledger technology rules. And
- 36:11now we have AI dispute clause and rules. And
- 36:15these largely track the commercial rules, but
- 36:18they have a few differences. First, they have
- 36:20a pre -done protective order that automatically
- 36:23applies because protective order issues are very
- 36:27frequent in AI -related disputes. And this avoids
- 36:30the cost and time and expense and challenges
- 36:32associated with negotiating a protective order.
- 36:35Second, it kind of follows a long standing debate
- 36:39about whether neutral should be generalists or
- 36:42subject matter experts and requires that neutrals
- 36:45in these cases be subject matter experts. So
- 36:49the benefit of that is that you are. having a
- 36:53neutral involved in a dispute involving something
- 36:55technical where the neutral has a basic technical
- 36:58understanding of things. And so you are not trying
- 37:02to explain digital technologies to someone working
- 37:05in an analog environment. These rules can be
- 37:11put directly into dispute resolution clauses,
- 37:14or they apply automatically when AI is at the
- 37:19heart of the substance of a dispute that comes
- 37:21to Jams. There's also a special process for reviewing
- 37:28highly confidential data. Often, AI -related
- 37:32disputes involve highly sensitive trade secrets
- 37:35and confidential information. training data sets
- 37:39and the like. And so there are enhanced protections
- 37:41around that. And there's also a provision for
- 37:45neutrals to appoint independent third party technical
- 37:48experts as opposed to traditional party experts
- 37:52if desired. All this in the mind of kind of having
- 37:56rules that are a bit more bespoke and solving
- 37:59some of the challenges that AI presents to dispute
- 38:02resolution, which can involve really highly technically
- 38:05challenging technologies, vast swaths of information,
- 38:11very contested discovery issues, and this is
- 38:14an effort to have ADR get the benefits associated
- 38:18with it faster, more efficient, and fairer dispute
- 38:21resolution. Other recent notable developments
- 38:27in the world of AI come from another DR provider
- 38:31or ADR provider, the American Arbitration Association.
- 38:34They have recently launched an AI arbitrator.
- 38:37So this is not rules about the substance of a
- 38:40dispute. This is very much on the process of
- 38:42dispute resolution. And at the moment, the AI
- 38:45arbitrator only applies to documentary evidence
- 38:49only construction disputes. which were selected
- 38:53because AI is not currently well -situated for
- 38:57resolving factual disputes, for example, evaluating
- 39:00witness credibility. It can do that. It is a
- 39:04controversial use case, but AAA is not using
- 39:07it for that. Construction disputes tend to be
- 39:10relatively simpler legally. They tend to have
- 39:13relatively more prescribed and applicable rules.
- 39:16So it is a good potential case study for this.
- 39:21And the way it works is both parties have to
- 39:23agree to use the AI arbitrator. You have an AI
- 39:27arbitrator that then appoints a human arbitrator.
- 39:30I just copied this rule because who would have
- 39:33thought one day we'd be looking at rules talking
- 39:35about how the AI arbitrator will select a single
- 39:38human arbitrator. There you have it. The AI arbitrator
- 39:41takes all party submissions, generates basically
- 39:44a list of claims from them and a statement of
- 39:46case, confirms with the parties. They're happy
- 39:49with that. It then gets submitted to the human
- 39:52neutral along with a draft award. And the human
- 39:55neutral, ultimately, the human arbitrator approves
- 39:59the draft award or can edit however they want
- 40:01the draft award. And so at the end of the day,
- 40:04it is a human award. So this is more sort of
- 40:08an AI augmented arbitrator. But why would someone
- 40:13do this? Well, the selling point is that AAA
- 40:17says it's going to be faster and cheaper. And
- 40:20this is certainly long but a selling point of
- 40:22using AI. It's kind of a selling point of anyone
- 40:24who's looking at automating something right now
- 40:26or augmenting. In theory, it will be cheaper
- 40:30and faster. And this is data that AAA promotes
- 40:34from kind of early investigation of it, not from
- 40:37having it used in actual practice. So we'll see
- 40:39what the data has. Some people also think that
- 40:43the AI outcomes may be fairer or better. To be
- 40:48sure, you know, that's a difficult sort of thing
- 40:51to measure because all of these disputes are,
- 40:55you know, fact dependent and there isn't a kind
- 40:59of correct objective benchmark for whether the
- 41:03AI is getting it right, though there are ways
- 41:05to try and measure something like this, you know,
- 41:08including, you know, in research by having panels
- 41:10of neutrals evaluate decisions and deciding if
- 41:12they're right. Some parties don't trust AI, some
- 41:18parties trust AI more than humans because they
- 41:20think they are unbiased or impartial, though
- 41:24the biases and partiality of the companies making
- 41:28AI have led to AI being called the fourth party
- 41:32in dispute resolution. And the AI company is
- 41:36the fifth party in dispute resolution. But in
- 41:39any case, while controversial, this has moved
- 41:41forward. So we now have AI -specific dispute
- 41:44resolution rules and AI arbitrators. And if you're
- 41:47curious at testing this out, it is available
- 41:50at a cost. But you can also run your documents
- 41:53through the AI arbitrator resolution simulator,
- 41:56which will give you kind of predictions on how
- 41:59an AI arbitrator might resolve the case. This
- 42:02is sort of like getting an early neutral evaluation
- 42:04of your case before taking it to mediation or
- 42:07arbitration. And dispute resolution, broadly
- 42:11speaking, AI is now and has been for a long time
- 42:15really autonomously resolving disputes. For any
- 42:18of you who use Uber or something like Uber, and
- 42:21these days that's probably all of you, very difficult
- 42:25to get a human being to resolve a dispute. Unless
- 42:28you say it's a safety issue and then explain
- 42:31that really it was that your promo code didn't
- 42:34work, if you try and initiate a dispute with
- 42:38Uber or these days half the companies out there,
- 42:41it will send you through an AI algorithm to just
- 42:43sort of resolve the thing, which can be both
- 42:47faster and cheaper and more effective or painfully
- 42:51frustrating, although so can be speaking to a
- 42:54customer service representative. The use of AI
- 42:58in dispute resolution is like the use of AI,
- 43:02well, like AI or the use of AI creatively. Not
- 43:04a new thing. People have been using it since
- 43:07the 70s. Insurance companies have been using
- 43:10it, for example, to help generate case outcomes
- 43:13and case value using good old fashioned AI, which
- 43:18was traditional programming. If victim is a minor,
- 43:22then increased settlement value by one third.
- 43:26If victim was careless, then decreased settlement
- 43:28value by one third. This is sort of programmed
- 43:31in based on expert feedback. Most of the modern
- 43:35AI systems people are very excited about are
- 43:37foundation models built by a handful of large
- 43:40companies like Google and OpenAI and Anthropic
- 43:44and Mistral. on which many other providers kind
- 43:48of specialize or wrap those models to do interesting
- 43:54things. And they're largely based on machine
- 43:57learning, and they're largely based on exposing
- 44:00AI systems to massive amounts of data. So very
- 44:03different way of programming it. Same basic outcome.
- 44:09AI's kind of first major success and debut in
- 44:13dispute resolution was with eBay in the early
- 44:162000s. So eBay, for those of you who remember
- 44:18using it, was a marketplace for buying and selling
- 44:21things. It still is. And early in the days of
- 44:24eBay, when there were disputes, parties would
- 44:27submit sort of open -ended complaints about what
- 44:30the dispute was, and a human neutral would try
- 44:32and get the parties to help resolve it. eBay
- 44:36set up an online dispute resolution process that
- 44:39was effectively AI driven. And they made a few
- 44:43key changes on this. One is they prevented parties
- 44:46from sending in a three page letter about why
- 44:49the item didn't match the seller's description.
- 44:52And they just put in a radio button that said,
- 44:54I bought something. I haven't received it. The
- 44:57item doesn't match the seller's description.
- 44:59And at least in initial phases of this, the system
- 45:02tries to resolve it for parties without the use
- 45:05of a human neutral. This had several appealing
- 45:09features as a case study. One, these were high
- 45:12volume, low value disputes. So if you're fighting
- 45:15over a $50 item, and my fee at Jams is somewhat
- 45:20in excess of $50 an hour, you couldn't really
- 45:24bring me in to mediate that dispute sensibly,
- 45:27or really anyone to bring in that dispute sensibly,
- 45:29right? So the low value of it made it sensible.
- 45:33It was high volume, also relatively simple. So
- 45:37eBay had some pretty simple rules on things.
- 45:39The risk of shipment is on the seller. Right?
- 45:42So buyer says they haven't gotten the item. Well,
- 45:44that's that, unless the seller can prove they've
- 45:46gotten the item. So you don't necessarily need
- 45:50a four -hour caucus to figure that out, something
- 45:54a system may help resolve. So eBay rolled this
- 45:57out and found that, to everyone's delight and
- 46:02surprise, not only did it vastly improve settlement
- 46:05rates, but it improved participant satisfaction.
- 46:09It turns out that people were happier having
- 46:12an automated system help them with a voluntary
- 46:15resolution process over kind of the neutral driven
- 46:18process. This quickly or eventually became the
- 46:23standard anywhere, internationally as well. It
- 46:26becomes extremely difficult these days to get
- 46:28human beings involved in some of these dispute
- 46:30resolutions except the kind of very high value
- 46:32levels. This sort of process is also making its
- 46:36way now into the courts. So in a civil court
- 46:40in New York now, you can register for online
- 46:43dispute resolution. There are not everyone can
- 46:47qualify. You have to be proficient with the use
- 46:49of computers. You have to watch a couple of videos.
- 46:51You can't have certain sorts of disabilities,
- 46:53for example. But once you're in there, the system
- 46:56has a automatic basically blind bidding process
- 47:01to try and get the parties to resolve a dispute
- 47:03before it even goes to the court. And this is
- 47:07something courts internationally, particularly
- 47:10in some jurisdictions like China, where they've
- 47:12been more aggressive with this, have been leaning
- 47:15into, particularly where the process remains
- 47:18voluntary and party -driven. There has not yet
- 47:22been much judicial adoption of binding, even
- 47:26reviewable decisions, although there have been
- 47:28in some jurisdictions, not in the US. There's
- 47:31also been for many years a variety of commercial
- 47:34products automating or augmenting kind of the
- 47:38neutrals role in dispute resolution You know,
- 47:40this is from one called smart settle Other than
- 47:44jams. I have no no affiliation with any of these
- 47:47of these products I mentioned, you know, but
- 47:49this allows parties to come in here and you know
- 47:52one party wants a million dollars and one wants
- 47:55to pay $300 ,000, and they put their starting
- 47:57bids in, and they put their bottom lines in,
- 48:00and the system helps them bridge the gap through
- 48:02an algorithm that isn't disclosed so that it
- 48:05can't be gamed. And there's benefits to a party
- 48:07for being willing to come down more earlier in
- 48:09terms of gap bridging. And in this particular
- 48:13system, for example, if the grid gap can't be
- 48:16entirely bridged, you have the option of going
- 48:18to a human neutral or referring it to that. These
- 48:22systems aren't just for people fighting over
- 48:24dollar amounts. There are many of these systems,
- 48:27they're used internationally and they even, you
- 48:30know, are used in things like child, you know,
- 48:33family dissolution. So involving things like
- 48:36child custody that aren't, you know, simple numerical
- 48:38figures being bridged. uh you know now we have
- 48:43kind of ai mediators this is a company called
- 48:45claim bright that is largely marketing itself
- 48:48to insurance companies and you know again the
- 48:51pitches this is an attempt to entirely replace
- 48:54me at a much lower price point than Jams. And
- 48:59so the mediator here is an AI bot and is based
- 49:04on, of course, as almost all these companies
- 49:06do, based on one of the foundation models, specialized
- 49:10to focus on mediation. And so these products
- 49:15are largely starting to roll out now and there
- 49:17is not yet good data on their outcomes. There's
- 49:21also training tools, for example, like Mediate,
- 49:24where if any of you want to try out your mediation
- 49:26skills, you can go on and practice with an AI
- 49:29bot and help the AI bot resolve some of its issues.
- 49:32It will help score you and give you feedback
- 49:34on that. You know, then also Liz mentioned kind
- 49:37of lawyerly guidelines. There are also neutral
- 49:40guidelines out there, you know, some proprietary
- 49:43to jams that are used internally, you know, some
- 49:45from the Silicon Valley Arbitration and Mediation
- 49:47Center, you know, dealing with, you know, how
- 49:50should neutrals and parties be using AI? And
- 49:54as Liz said, you can't get away from this now.
- 49:58Everyone is using AI. Some of the times when
- 50:02I mediate, I'll go to a room, ask someone something,
- 50:05they'll ask ChatGDP, maybe their lawyer, maybe
- 50:07ChatGDP. I'll go to the other room, they'll ask
- 50:09ChatGDP, maybe their lawyer. So parties are using
- 50:13it. Counselors are using it to write mediation
- 50:15briefs and to kind of give predictions on case
- 50:18value. and to create profiles of arbitrators
- 50:22and mediators. So these are being used extensively
- 50:24as well. And neutrals are using them, both for
- 50:28guidance and mediation and for assistance in
- 50:31arbitration. And there are not yet very clear
- 50:34guidelines on what's permissible and what's not.
- 50:37I think the keys are one, confidentiality. So
- 50:41you have to be very aware of whether the AI system
- 50:45you're using maintains everything you provided
- 50:48in confidence. If you're using the free version
- 50:51of ChatGPT, for example, it does not treat what
- 50:54you put in confidentially. And so there was an
- 50:57interesting case a couple of weeks ago where
- 50:58a court found a client had waived privilege because
- 51:01they basically asked ChatGPT for legal advice.
- 51:05CHTTP says, A, this isn't confidential, then
- 51:08B, we don't give you legal advice, so don't use
- 51:10it for that. Of course, people can't be helped.
- 51:14If, as a neutral, you were doing that too, a
- 51:17month from now, someone might ask it a query,
- 51:19and it might say, well, in this dispute you had
- 51:22with Nova Nordisk, they offered $1 million as
- 51:25a starting point, and that would be bad. Or it
- 51:28could be susceptible to discovery. A little less
- 51:32clear is transparency and disclosure to parties.
- 51:35When is a disclosure required? It's always safer
- 51:39to disclose and get permission from the use of
- 51:42things. It could well be possible. The parties
- 51:44could dispute, for example, an arbitration award
- 51:47if a neutral used AI without disclosing it. I
- 51:50think in that sense, it's not too unlike best
- 51:52practices for using clerks. And I think, again,
- 51:55the best practice is always to disclose things.
- 51:58And there are risks, for example, if the parties
- 52:00are giving you 500 documents they want you to
- 52:02review with dumping those all in an AI and asking
- 52:05it to summarize those documents for you. At least
- 52:09at the higher levels of things, the parties are
- 52:11bringing you on to exercise your judgment, unless
- 52:15you're telling them you want to use a clerk or
- 52:17you're telling them you want to use AI, and they
- 52:19approve of that. I'll end with this. AI is getting
- 52:25better and people aren't getting equivalently
- 52:28better. And so we have seen really some dramatic
- 52:31improvements per Will Smith in the past few years.
- 52:34It's coming faster and faster month by month.
- 52:37And this is a chart from a book by Ray Kurzweil
- 52:40from about 20 years ago, which shows basically
- 52:43this upward trajectory on AI progress. Sam Altman
- 52:46and Elon Musk say in the next three to five years
- 52:49at most, we'll have AI that can broadly do any
- 52:52intellectual task a person can do. Elon Musk
- 52:56said we'd all have self -driving cars a decade
- 52:58ago. So I take it with a grain of salt. He says
- 53:01a lot of things. I do have a self -driving Tesla,
- 53:03though. And I do like it more than I did seven
- 53:06years ago when it used to Phantom Brake on the
- 53:08freeway all the time. But anyway. Whether we
- 53:13get there or not, the next prediction is, well,
- 53:16then these computers will start improving themselves
- 53:18and become super intelligent. And there'll be
- 53:20really nothing for human beings left to do anymore.
- 53:23These are controversial claims and I make no
- 53:27personal claim to when and if we're gonna get
- 53:29artificial in general intelligence and artificial
- 53:31super intelligence. But the point is valid. These
- 53:34systems are only going to get better. So we will
- 53:37only get more adoption of these things. We will
- 53:40only get more sophisticated systems and there
- 53:42is no getting away from it. So now is the time
- 53:45to figure it out for yourself. Alright, I'll
- 53:49stop there. So we have a bit of questions and
- 53:51it looks like we have some questions and comments
- 53:53unless these are Disclaiming my my offer from
- 53:57the Bar Association to buy you all all a book
- 54:02Liz you want to start with these or I think that
- 54:10they may be mostly for you the Around and I think
- 54:14you might have covered some of them Right. The
- 54:17first question is, I am a mediator. What duties
- 54:19of mediator to use AI for case analysis? For
- 54:22example, I have developed a program to do analysis
- 54:24of claims. Any suggestions or warnings using
- 54:27research and analysis using Claude work? Yeah.
- 54:31So I mean, I do think it's different in mediation
- 54:33versus arbitration. I mean, there's less in mediation.
- 54:37If you want to use AI to help you better understand
- 54:42the claims of the parties or to be a bit more
- 54:44evaluative, I don't think there's any firm rule
- 54:48against that. I do think it is kind of a safe
- 54:52practice and possibly a best practice to disclose
- 54:55to clients that you use generative AI. Again,
- 54:58to make sure you're using something with an enterprise
- 55:00version that's keeping it confidential. And to
- 55:06the extent it's just helping you to understand
- 55:08things, I think that that is fine. Bearing in
- 55:11mind also AI isn't perfect. Aside from phantom
- 55:14breaking in the freeway, it will hallucinate
- 55:16things, whether it's cases or it won't accurately
- 55:18reflect the statements of the parties. So I always
- 55:22trust but verify and kind of make sure I've done
- 55:25enough diligence to be comfortable the AI isn't
- 55:27just making something up. The next question was,
- 55:31what platform are you using to generate the image
- 55:33of the cat cooking? And what platform can you
- 55:36use to generate videos based on a prompt? Ah,
- 55:39well, this is perhaps a little more technical
- 55:41than I was here to answer. But I used Gemini
- 55:44for that. And if you used, so I picked Gemini
- 55:48for no particular reason other than I wanted
- 55:50to pay for an enterprise version of something
- 55:52and I got Gemini early because I vaguely trust
- 55:55Google compared to alternatives and have sort
- 55:57of stuck with that. If you prompt in Gemini,
- 56:00it will direct you to their video platforms and
- 56:03help you make videos. But there are many free
- 56:06platforms for entering a prompt and getting a
- 56:09video, like make me a video of Will Smith eating
- 56:12spaghetti. I think if you just use the free ChatGTP
- 56:15version, it will do that or link you to it. It's
- 56:20not too hard to find online. We have a third
- 56:25one, are Mr. Abbott's slides available? I don't
- 56:28think so. They're pretty much just pictures anyway.
- 56:31But not to worry, for those of you who want to
- 56:33read not quite an encyclopedia, Liz and I have
- 56:36written several things. And the Beverly Hills
- 56:38Bar Association is sending them out to all of
- 56:41you. You are welcome to have AI summarize them
- 56:44for you. And then we had here, I would like to
- 56:47confirm that at this time, AI is not considered
- 56:49confidential because it's from a third party.
- 56:52I think that that is a fair point, but not entirely
- 56:57correct. I think it is a complicated answer to
- 57:00that question. In the facts of the case that
- 57:02I mentioned, for example, a court found that
- 57:05a communication was not confidential or privileged
- 57:09because it was to a third party platform under
- 57:12a term of service where the communication was
- 57:16not confidential, privileged, or attorney -client
- 57:18privilege. If I was communicating now in a mediation
- 57:25over Zoom and the Zoom was being recorded on
- 57:28the Zoom platform, it would be a third party
- 57:31hosting data. But there is an agreement in place
- 57:35for Zoom to treat that information as confidential.
- 57:38So I think the lawyer answer, it depends. You
- 57:41have to look at your agreement with the third
- 57:43party in the context of the third party. If you're
- 57:47using it under an enterprise license the way
- 57:49that you would use Excel or Word or Outlook,
- 57:53I don't think you will have. significant confidentiality
- 57:57concerns, but it does depend. And when will the
- 58:00Beverly Hills Bar send out the book you referenced,
- 58:03Ryan Abbott? Yes, I think someone is going to
- 58:06get in a van right now. If you just send them
- 58:08your name and address, they're just going to
- 58:09kind of like Paperboys used to do back in the
- 58:12day, just toss it toward your office. Should
- 58:14we do a full disclosure? Full disclosure, that
- 58:17was humorous. You're not really getting this
- 58:19book. I thought an encyclopedia would be larger,
- 58:23though. It is pretty small. So it doesn't even
- 58:25work well as a door jam. But it doesn't cause
- 58:29back pain lifting it. Those are true. And I would
- 58:32agree on the answer about confidentiality, that
- 58:34it really depends on the platform you're using,
- 58:37your agreement with that platform, their security
- 58:39procedures, and if they're retaining the data.
- 58:44And that's something that is. needs a lot of
- 58:47due diligence. So public -facing AI platforms,
- 58:49though, there's a lot more risk involved. We
- 58:54have 60 seconds if anyone else has a burning
- 58:56question, or if anyone's Otter AI note -taker
- 58:59would like to chime in with a question. Oh, here
- 59:00we got one more. Is it helpful to use multiple
- 59:05AI apps to compare answers for issues of mediation,
- 59:08getting different perspectives such as Claude,
- 59:10chat, GTP, and perplexity? Well, that is certainly
- 59:15something people do. If you use different platforms
- 59:21and give it the same prom, you may get a substantially
- 59:24different answer to it. As a mediator myself,
- 59:30I'm not sure that I personally need that many
- 59:33different opinions from AI systems on some issue
- 59:36that I'm facing. Um, but you know, I mean, that
- 59:39is a problem solving technique and for certainly
- 59:41for critical analysis or important questions,
- 59:44people do do it. You know, the cost of that goes
- 59:47up if you're using enterprise versions of things
- 59:49that have multiple subscriptions. Um, but, but
- 59:53I mean, you are welcome to play around with it
- 59:55and kind of see if you get better answers from
- 59:57one of the other, you know, people, you know,
- 1:00:01it's hit or miss what you get from any one system
- 1:00:03at any point in time. I just wanted to jump in
- 1:00:09to address a prior question that was asked. And
- 1:00:11I think I answered it while I was speaking, but
- 1:00:13that the data could be outdated from studies
- 1:00:16about AI. And that is, yes, the data is outdated
- 1:00:21almost immediately from studies about AI because
- 1:00:23everything is moving so quickly. And that's why
- 1:00:26I put that caveat in the presentation as well,
- 1:00:28that take the numbers with a grain of salt, but
- 1:00:30definitely they're. where there was marketing
- 1:00:34coming from companies about products being hallucination
- 1:00:37-free, there were hallucinations in the products
- 1:00:40then. So just do your own due diligence on any
- 1:00:43tool that you're using. On that note, I think
- 1:00:47that we have reached the top of the hour. So
- 1:00:49for the 58 out of 59 of you who were here for
- 1:00:53your CLE credits, I think that you have been
- 1:00:56punished enough. and go forth and have fun with
- 1:00:59AI. And if you want to reach either Liz or myself
- 1:01:02going forward, we're easy to find online. All
- 1:01:08righty. This concludes today's presentation.
- 1:01:09Thank you so much, Ryan and Elizabeth, for an
- 1:01:11excellent presentation. And thank you so much
- 1:01:13to our audience. Have a great day, everybody.
- 1:01:17Thank you.