Welcome, everyone, to today's presentation of AI - ChatGPT Attorney Ethics. My name is Cari Sheehan and I will be your presenter today. Just a little bit about myself before we get started. I'm an assistant clinical professor of business law and ethics, and I'm also a conflicts attorney in traditional legal practice. As a conflict attorney, I deal with ethics all day, every day, and I also teach it as a professor. So you can say attorney ethics is my bread and butter, my 24/7. And I just happened to start speaking on the ChatGPT and the artificial intelligence because as a professor and as a practicing attorney, it has implications in both worlds. As a professor, I'm looking at it from are my students going to use it to cheat? Are they going to use it to try to not do their work like they're supposed to? And then from the legal standpoint of it, too, in the legal world, we look at it from how is it going to affect us ethically? Is it going to breach our client confidences? Can we use it? Should we use it? Is it accurate? Is it not? And so a lot of those things that I just mentioned are what we're going to talk about here today and hopefully, you know, get some insight, get some good understanding and some good tips just for your practice and maybe your firm or where you're working for is developing their own policy or their own ChatGPT policy. Maybe they're not, but it is something to think about, especially if you're thinking of using it or if you already are using it in your practice. So let's get started. First, I'm going to just talk about a little overview of ChatGPT and then about the ethical concerns that we see. And then today we're also going to try it out together. I'm going to pull it up on the screen. We're going to look through it. We're going to see, you know, the ins and outs and what it has to offer. And two, we're going to test the accuracy of it. And that can always be semi exciting when we test the accuracy of ChatGPT. So without further ado, ChatGPT, if you do not know what it stands for, it stands for Generative Pre-trained Transformer. So ChatGPT is not like a human brain. Yes, it does give us responses and human context and human verbiage. Like you're talking to someone just in a casual conversation. But it's a pre-trained transformer, which means it's only as good as its training or it's only as good as the information input to it. And it takes that information and tries to transform it into a human response based on whatever query or. Q you input into it. It tries to then generate it out into a human response at that point. Now, where does it get its information? Well, it gets its information from several different sources. The main place it gets its information, though, is the Internet and then whatever else. The creators, which is OpenAI is the creator, input into it. It can also and that stuff could be from books, from different articles, from different prompts, from people who are using it, which is a concern we'll see here in a few minutes. But anything that is programmed into it, input into it or connected to the Internet is really where ChatGPT draws its information to generate out responses, which is kind of amazing because if you think about it, when you go into Google, you may enter query, say, you know, what is the top ten list of this? And Google is going to give you different Web pages to maybe look at, maybe give you a quick response or something like that. But you still have to link out and read those web pages. You still have to link out and try to put the information into like a human form. If you're trying to write a paper or if you're trying to do something like that. But what ChatGPT does, it takes that second step out of it. It's kind of like when you enter your query, just like you would on a Google ChatGPT does all the scanning of all those internet pages that would come back, summarizes it for you and gives you a response like you would in a human context. So it takes out almost that whole research compiling reading multiple sources, step from just a normal Google search, which is pretty amazing. But is it that great? And should we hang our hats on it and think this is the next greatest thing? And that's what we've been trying to figure out. Chatgpt was actually released back in November 2022, and since then, it's gone through different models. We're actually at ChatGPT four, so the fourth generation, which can do more things and each time it does get more accurate and it does get more like, Oh, we can really rely on this, but should we rely on it as lawyers? And that's what we're going to really delve into. But just some other things that ChatGPT can do, just so you know, kind of the breadth breadth of it. It has a language generation feature. It can generate text in natural language, and it's difficult to distinguish it from a human being. It's difficult to say. Did a machine write that or did a human being write that? And so when you do use it and when we start to run some searches, some queries here together, you're going to see that and you're going to think, well, how do I catch this? And as a professor, that's one of the biggest questions I ask myself. And as a lawyer. And even if you were a judicial a judicial judge or something, that's another question, too. How do you know this is the lawyer's work product? How do you know this is what it's supposed to be? It can also answer just general questions on any given topic. You can enter in there. Give me all the NFL teams across the United States. You can enter in there something else in those queries and then it's going to generate it back and you can find it. It can also complete text. So if you have a sentence or something that you just don't know how to finish it or finish a paragraph, you can put the first part of it in there and then query it up to finish out your paragraph or finish out your sentence. And it would do that so it can complete that partially written text and just any contextual relevancy there too. Some other things that ChatGPT can do. It can summarize text. It can summarize different articles. It can summarize information. Like if you put into it just a bunch of data and you say, give me this in a summary paragraph. It can make it very short, concise, condensed version of it. So that can be very helpful as well. It can also translate language for you. It can be kind of an interpreter in that context. So say for instance, you have a client or you are talking to someone and they don't speak English. You can type in what they're saying or you can type in what you're saying and say, translate it to this and it will translate it and then the person can read it on the screen. So it's almost like a built in translator right there at your desktop, which is a nice feature of it too. Now again, we're going to look at the accuracy of that and if there are any ups and downs to that as well. It can analyze sometimes tone and sentiment and text, but it's not always great at that. Remember, this is a pre-trained transformer, and so that means it cannot feel emotions like a human being. It can't draw on past experiences when it gives its responses. It can't draw on situational context because it doesn't know the situation. It just knows the data that you put into it and the data it's drawing from. It's different sources. And so it's lacking those things that as human beings we use when we do skills such as critical thinking, critical reasoning, or just saying, you know, I've been here, I've done this in the past, even though the textbooks say this, this is actually what's really going to happen, this being pre-trained, it's going to give you the textbook response every time. And we all know from practice and just from life, sometimes the textbook response is not the best response to have. So we really have to be very careful with when we use this and how we use it, because we have to remember why we think it is a human being. It is not a human being. So we really need to think about that and we really need to be careful as we'll see as we go through the different ethical rules. And it can also the last one here on the screen chat with you. If you're bored at home and want to talk with someone, it can hold a conversation. You can type in questions or just conversation, and it's going to type something back like a human response. The thing about Chat GPT is it is trained to always produce a response no matter what. And so even if it does not have the accurate information or the most up to date information, or if it doesn't know at all, it's going to make up something to give it to you because that is what it is trained to do. And so you really need to be careful about that too, because you may think, well, that's kind of neat. You know, chatting is going to make up a conversation. But if you're doing it for more like a legal research or you're doing it for more of a factual type basis where you want to learn something factually, that can be a really scary thought because if it's going to generate a response, whether accurate or not accurate, whether it knows the answer, it doesn't know the answer. How do you know which one it is? It's not going to come out and say, I have no idea I'm lying to you right now or I have no idea. But here's just a guess. It's not going to say that. So you really have to be careful. And we'll see when we get to the live version of Chat GPT Here at the end of the seminar on the main screen, one of the main limitations is that accuracy and it does not guarantee accurate results. So we need to be careful while it has all these great features and it is revolutionizing the way humans interact and the way we use the internet and just our capabilities and what we can and cannot do and not just chat GPT in that context, but just AI in general artificial intelligence. It is coming at us mainstream and it is something to kind of to put into a little bit of context. In the first months that GPT hit the market, there was over 13 million users to the program. Now that is more than TikTok. When TikTok launched, it took TikTok probably about 6 to 9 months to hit that many users on that grand of a scale. But chat GPT hit it on a monthly basis in the first month of coming out on the market. So that tells you anything and how much humans and people are really into this technology and how much they feel like it's going to shape their lives, make it easier, make customer service and content creation a lot easier for them. They are on board for it, the good and the bad. But we really need to look out for the bad. And just while the general public thinks it's this end all, be all. And it is amazing. And even like businesses, the general public and normal businesses are not subject to the rules of professional conduct. That means they are not governed by the same standards we as lawyers are governed by. They did not raise their right hand to take that oath, to uphold the Constitution and abide by the, you know, the laws of the jurisdiction and our rules. They just didn't. And so while as general citizens, we can use it maybe for general things, we really need to take a hard look at ChatGPT to see if it complies with the rules of professional conduct. Because when we raised our right hand to become lawyers, we took an oath that we would uphold those standards and those standards become. Before anything else when it comes to our law license. And those standards are the rules of professional conduct. We have to remember, apply 24 hours a day, seven days a week, and our professional and our personal capacity. So we really need to look at this technology and see where the flaws are and see where us as lawyers really need to step back and say, okay, wait a second, while this is revolutionizing the way people work and think and create stuff, is it okay for me as a lawyer to use it? The first rule I do want to talk about in this is rule one, point one competency model. Rule 1.1 tells us that lawyers must provide competent representation to a client. This means that we have to have the requisite skill, knowledge, thoroughness, preparedness to effectively represent our clients. And this does not just mean that we have to know the law. For instance, like if you are a medical malpractice attorney, it doesn't just mean you need to know how to file a medical malpractice lawsuit. It means more than that. It means that you need to be knowledgeable in anything that touches and concerns the profession, which includes technology, technology touches and concerns the profession, whether we like it or not. And it's comment six of Rule 1.1 that says to maintain the requisite knowledge and skill, a lawyer should keep abreast of changes in the law and its practice, including the benefits and risks associated with technology relevant to the lawyer's practice. So if we just stop right there, and if you think about it, ChatGPT an artificial intelligence in general is associated and relevant to a lawyer's practice, and it's associated irrelevant whether you use it or not because clients are using it. Maybe opposing counsel is using it, maybe opposing parties are using it. Are your vendors and experts using it? Are there other people you deal with using it? Do you know? Could you definitively answer the question as to if your legal administrative assistants and your secretaries and your non-attorney staff are using it or not in producing work product to give to you? That's a scary thought. How many of you could sit there and definitively say, No, my staff doesn't use that? How do you know? Are you watching them on the computer the whole time? Have you trained them in the process? Have you thought about what that would look like? Have you talked to them say, hey, we shouldn't use this yet? Or do they know about it from their kids or from, you know, themselves researching into it? And when you ask them to draft a letter or you ask them to tee up interrogatories, maybe they want to take the easy way out. So they put it into ChatGPT and chat. Gpt does it for them and then they give it to you. Can you definitively say that has never happened in your firm? Most of you will say, Well, yeah, I can, but then you start thinking about it. You're like, Well, I'm not watching them every second. I'm not doing that every second. We don't have a policy on it. It's a scary thought and it's a scary thought because you are responsible for the actions of your non-attorney assistants during the practice of law. So if they violate the rules of professional conduct, you violate the rules of professional conduct because they don't hold a law license and the disciplinary commission is going to come after you. If you task your paralegal, your assistant with drafting a brief or putting in case citations and they use ChatGPT and they gave an inaccurate result and then they gave that to you and you didn't double check it, who's going to be in trouble if that goes to the court? And then you're lying to the tribunal or presenting false information. You are not your assistant that drafted it. You are as their supervisor. So we really need to stay competent in this. Even if you decide you don't want to use it yourself, you need to stay competent so you can educate your staff too, and educate your clients who are probably going to be using it as well. To know the risks and benefits of it. Some other things to considerations. Chatgpt does not tell us where it gets its information. So say, for instance, I query up a thing that says Give me the top ten cases regarding affirmative action or give me the top ten cases regarding X, y, z. It will give me ten cases. Now, are they going to be the top ten? Maybe not. Are they going to be ten cases? Yes. Are they going to be citations with them? Normally, if I queue it up right. Are those citations going to be accurate? Yeah, I maybe give it 30% accuracy based on what I've tried and input into it before. But the thing is, it's not telling me. I drew these cases from Westlaw. I drew these cases from LexisNexis. I took this from this brief I found on the Internet. It does not show me where it gets the information it's producing to me, which can be a problem because then how do we verify the information? How do we verify the accuracy without double checking it somewhere else? So that is one of the things you need to keep in mind and consider when using ChatGPT is if someone says, Where did you get this case citation, you're like ChatGPT or ChatGPT get it? I have no idea. We can't answer that. So we really need to think about that too. Chatgpt does not does not guarantee the accuracy or completeness of the information it provides which can lead to disciplinary things. There was actually a case recently out of New York, which is there in the next bullet point, a sub bullet point of it, where a lawyer asked ChatGPT to draft a brief or give citations for a brief and then use what ChatGPT said as gold put it in the brief, it actually went through another partner, another counsel, because the one who was drafting it was not licensed in the court in the jurisdiction. So they were using a local counsel situation. They both signed off on it. It went to the judge. Opposing counsel got it because they were going to respond to it. They went to look up the cases to do an adequate, accurate response and be comprehensive. And they couldn't find the cases anywhere. They asked opposing counsel, where do you get these cases or these citations wrong? You know, we're not finding these cases. Opposing counsel's only argument was, well, we got it from ChatGPT because they couldn't find it either. Chatgpt didn't tell them where they found the cases. Chatgpt probably didn't have cases on point with the argument that these attorneys were wanting, but it is programmed to provide an answer whether accurate or not. So it gave an inaccurate response and these attorneys blindly relied upon it, turned it into a court, were up for sanctions this past June of 2023 and got sanctioned $5,000 for the conduct. Now, I have not seen any disciplinary commission violations come down yet in that case, but they're usually slower moving and sometimes those settle and you never hear about them. But that was the first case that has gone across the nation, really, where a lawyer has used ChatGPT to its detriment. And it really highlights the inaccuracies that can be found in it, the false results, the false cases, the false data, and that we cannot verify where it gets it from. Now, what that lawyer should have done to be prudent is take the citations that it was provided by ChatGPT, put them into Lexis, Westlaw, or whatever search engine they're using to do legal research and double check them, made sure they were accurate, actually gone out and read the cases to make sure it was what they were saying it was. But the lawyers didn't take that step. And so staying competent, knowing that ChatGPT has these flaws, especially this inaccuracy and incompleteness is vitally important, especially if you're going to be using it for your law practice. Another thing too, to remember just on competency reasons is ChatGPT is limited to contents and events up to 2021, and anything after that, it does not guarantee it will pick up. So again, it's a pre-trained transformer, so anything input in it, anything it can pull is what it's going to summarize and give you. If the data is only guaranteed up to 2021 and that's not even guaranteed accurate, then that's another limitation of it too. And that's a huge problem. Think of the groundbreaking cases that just came out from the Supreme Court. Think of the law that's been created from 2021 till now in 2023. Think of all those things and ChatGPT is probably not pulling from West Lawn, LexisNexis and all those great legal research search engines because those are passwords. Those. Our subscription services. It's pulling from things that are more free services that are more out there to the general public or that they have a contract with, you know, a database or something like that where the programmers of ChatGPT have entered the data in. So you've really got to think that the content is limited, especially in the legal context. It's not guaranteed to be accurate. We've already seen our first lawyer sanctioned for it out of New York, and it doesn't even inform you where it gets this information. These are all things to know because if you didn't know these things and just used it, you're going to be that next test case. Like the lawyer in New York probably getting sanctioned for using it and not being accurate or having other issues. Another thing to consider with using ChatGPT is just data privacy and confidentiality in general. We all know or should know that we have a duty to keep our client confidences secret, which means we cannot reveal any information relating to the representation of a client unless the client gives informed consent confirmed in writing. We also have to take reasonable steps to prevent disclosure of confidential information and making sure it does not go out to third parties where we don't want it to. With this, if you think about it, if you want ChatGPT to generate a brief, you want it to generate a derogatory or a letter to your client, what do you have to do? You have to query it up to do that. So you have to feed ChatGPT information probably about your client, including your client's name, details about the case, facts about the case, so that it can then create or queue up what you have asked it to. If you don't do that or give it those specific information, it's going to give you a general response to a general question that's probably not going to be relevant to anything you need and definitely couldn't create a document you need if it doesn't have the specifics. The problem with that is when you put a query or information into ChatGPT, it stores it. Now, there are some features that I'm learning about where you can say, okay, I don't want you to store it forever, but it still stores it for at least 30 days. And that's only if you have an account or you turn on this button or, you know, some limited thing to only store it for 30 days, but it still stores it. And that's a problem because when it stores the information, it's acting like a little black box and the little black box retains all this stuff. And remember, I said it's a pre-trained transformer that draws from information fed into it, and then it can draw from the Internet. It's going to draw from the cues and queries you feed into. It is now in its little black box of pre-trained knowledge and pre-trained information. And then it's going to transform the information you put into it to generate out responses to another person's query if it believes it is relevant so it can actually use and store your client confidential data you put into it to generate out responses to other people. This is a huge or potential huge violation of Rule 1.6 confidentiality, especially if you put the client's name in there. Anything about the client's case, any facts, any details, anything at all? We're going to have a 1.6 violation. And that is just because that's how it works. It's a pre-trained transformer. It's like a black box that just sucks in all this information and then uses it to formulate answers to other people's questions. So you may think, well, they're not going to know it's from me or they're not going to know it's from that. That's probably true, but it's still you've still violated 1.6. And so you really need to keep that confidential information and watch out for that. And then to chat. Chatgpt is subject to being hacked, too. So even if they don't generate your data out to another person, and even if you could prove all that, it is subject to being hacked as well. And it was actually in March of 2023 chat GPT was hacked. It experienced a data leak through its open source library. By allowing users to see chat history of other users. So any chat history or anything you put into it was able to be seen by other users. That's a breach. In addition to that, it used open source libraries, which, if you don't know what an open source library is, it's basically an interface that stores and makes information readily accessible and it can configure it, it can document it, it can make templates codes, but it stores information to make it readily accessible. Well, the thing was ChatGPT used these open source libraries not only to store the queries which could be seen by other people, but also to collect user information for faster recon accessing. So in the open source, libraries were breached. The people who breached it, the cyber criminals were getting information on the users. They were getting the users names, they were getting credit card information. The last four digits, email addresses, phone numbers, different things like that were getting leaked out of these open source libraries. Now, this only affected about 1% of ChatGPT users, but that's only because ChatGPT is a free program. You only pay for a subscription if you want to be able to access it even in high volume times, meaning the platform itself can only support so many users at a time. So if it's a high volume time, you may not be able to get on to it with your free access or it's going to be slower. If you pay for a subscription, then you're guaranteed access at any time you take priority over the people who don't pay. So this breach, while it did release this information of the subscribing members of ChatGPT, it was only about 1% of those using it, but that 1% was still an impact and it was still releasing their sensitive information out there. So we really have to be careful because that's, I mean, maybe personal to the user. But think about if that was your client information. Maybe you put your client's payment information in there, their date of birth, their Social Security number, Whatever you put in there, it could be leaked, it could be hacked. It could be used to respond to other people's queries. You have no control once it's in there and. Question and cue it up. Chatgpt can use it, regenerate it, re transform it out to another user to see, you know, to answer different questions. So you've got to be very, very careful when you enter stuff into the database or when you enter in stuff into ChatGPT because otherwise you don't know where it's going. And you could have a huge violation of Rule 1.6. And it was kind of interesting. I actually queued up into ChatGPT. I ask it, do you comply with the rules of ethics 1.6 regarding confidentiality? Because I want to know if it thought it complied with confidentiality for lawyers. And the response that ChatGPT gave me wasn't any shock to me. But it is just kind of interesting to note It said as an AI language model, I do not have the ability to practice law or provide legal advice and I am not subject to the American Bar Association's model. Rules of professional conduct, including rule 1.6 confidentiality of information. So it's telling me it doesn't have to keep information confidential. It's not subject to that. It's not a lawyer. And that's true. It's not this was not created for lawyers. This was created for the general public, us lawyers. We have to abide by the rules of professional conduct. Chatgpt went on to say, However, as an AI language model, I am designed to protect the privacy and confidentiality of users information and data. I use various techniques to ensure the data and information are kept secure and confidential. Nonetheless, it is important for users to remember that they are responsible for protecting their own confidential information and should use discretion when sharing stuff online. So it says, yes, we know we need to protect user information, but ultimately it's your own responsibility. So it's our own responsibility as lawyers to know what we put in there could be generated out, that we need to maintain our confidences with our clients and just the risk associated with ChatGPT in general. It is huge to remember that and to think about it, because otherwise we're going to run afoul with that rule. Some other rules of ethics that are worth noting that play into ChatGPT and artificial intelligence, a communication and accountability with clients. And this is one that a lot of attorneys forget because they think, okay, I'm the boss, you know, I'm the lawyer. I'm the one telling the client, you know, how we're going to proceed with the case. What's the best route that we need? This expert that I need to do this research, But we forget rule 1.2, which is not on your screen, but the rules of professional conduct 1.2 says and mandates that we always remember that the client is the boss. That is the number one rule to always remember as a lawyer that the client is the boss. What's the second rule in the legal profession? Never forget rule number one. The client is the boss. We have to consult with the client. We have to inform the client of decisions we make in the case. We have to seek their permission to use certain things. If you're going to use less Lexis and Westlaw and charge the client for it, you need to let them know that maybe they don't want you using those. Maybe they don't want to incur that cost. We have to keep them apprised of what we find, what we do, how we do it. You know what? We're filing different things like that. And at any time the client can say no, the client can say no, I don't want you to use Westlaw. No, I don't want you to use LexisNexis. No, I don't want you to use ChatGPT. That is the next conversation we're going to have to have with your clients. Right now, I know a lot of big law firms and even small law firms still seek permission to correspond with the client via email just because of the risk associated with an email being intercepted or hacked, or just the vulnerability of email accounts in general. So a lot of attorneys will consult with the client and say, Hey, I like to communicate by email. Do you approve of this? Here are the risks associated with it. The client says yes, they'll communicate by email. If the client says no, I prefer to keep everything in snail mail or letter form portion or excuse me, letter creation or by telephone verbal. Then you have to comply with that. Chatgpt is no different. If you have a conversation with a client and the client says, I don't want you using ChatGPT, I've heard bad things about it, or what am I paying you for? If you're using ChatGPT, can't I go out and use that? Isn't it a free source for me as well? And if the client says no, the answer is no. But we have a duty as the lawyer to keep the client reasonably informed of the status of the matter, the means of which we are using to accomplish the client's objectives. And those means, if they include ChatGPT, we need to let the client know that and we need to make sure they understand the risk and benefits of it and approve it. Now the problem with this is and you can just think in your head, you're like, okay, I'll just add it to my engagement letter. Okay, That's an easy conversation with my sophisticated clients. Okay? You know, we've got all this, you know, we can do, but just think about how this really plays out. You sit down with a client, you know, maybe you're talking to them verbally, maybe a zoom, maybe your cinnamon email or a letter or as part of the initial consultation, wherever you're at and talking with the client. And you sit down and say, Hey, have you ever heard of ChatGPT? They say, yes, we have, you know, used it for so and so or no, I haven't never used it. So then you explain to them what it is, if they've never used it, and then you say, I want to use ChatGPT to help me draft my legal documents. Think about if you were the client, if you were the client sitting there and someone said that to you, and then they you look at them or the client looks at you and it's like, what am I paying you the lawyer for? If it was me and I was a client and my lawyer said that to me, I'd be like, I'm paying you because of your brain, your knowledge, your skills. I think you're the best at what you do. So why are you using ChatGPT, which is not your brain, your knowledge, your skills? It's a it's an AI pre-trained transformer giving you something and you're not doing the work yourself. All you're doing is answering a question To me, If I was the client that would give me pause. I'd be like, Well, why am I paying you? I can do that myself. Why am I trying to get your expertise when you're just using an artificial intelligence pre-trained transformer? Are you not intelligent as a lawyer? Just think of those different questions and then to if you're going to use it, how are you going to charge for it? It's not like Westlaw or Lexis where you know you can bill by the hour or Bill by the time bill by the search, and it will generate out those reports for you to attach to an invoice with a client. This is a free service. And even if you bought the subscription, how do you divide that subscription up among all your clients you'd use it for? Then also too, if I come to you and say I'm going to use ChatGPT and I enter in my cue, I enter in the information I've advised you of the 1.6 risk. I've advised you of the data leakage and all of those things risk and you still as the client consent to it. But then I enter in the. It generates out the brief. I copy paste it. I'm done with in five minutes if I work on the billable hour. I just really kept my bills in half of my income or maybe more than half, because normally drafting a brief takes more than five minutes. Doing the research, doing the writing, doing the editing, doing the work behind it, you've got at least probably maybe a 1 to 3, you know, 4 or 5 billable hours, depending on how big you're writing. Even with the letter, you can have a 0.2, a 0.52 draft, a nice, you know, hearty letter. Chatgpt can do it in a minute or less. You've just cut your billable time. You need to think about that because you cannot. And we'll see this in the next slides when we really get into the fee section, Bill, for what you would have spent on it without GPT. You have to bill for what you actually spent on it. So while this may seem cost effective or time effective in drafting stuff, is it the best option? And do you really want to have that conversation with your clients and get their consent? Do you want to tell the client, Hey, I'm using chat GPT. If the answer is no, you can't use it. You need to get the client consent and communicate that you are using this with the client so they're aware of it. They also need to be aware of it for when they see a bill and it says 0.1 for a brief. They're like, How can you write a brief in 0.1? Well, I use chat GPT. Think about that conversation. You really need to think about those things if you're going to use a bit. You need to communicate with the client. You need to a minimum communicate if you're going to be entering in confidential information because we could violate rule 1.6. And the only way we can give over confidential information is if the client consents. So that communication with the client is huge, huge. Moving forward, knowing the risk and benefits. Going back into more a little bit about the fees, Rule 1.5 is our fee rule. It is instructive regarding fees, but it is not definitive and there is no comment. There is no section of rule 1.5 that is in regards to chat GPT and technology. It only outlines for lawyers certain things that would make a fee reasonable, and that's kind of the standard. Our fees have to be reasonable. So that's where it comes into not overcharging clients for the time saved or spent using chat. Gpt You should only bill for your actual time spent, not for what you would have spent absent that new technology. So that's something to think of if you're a plaintiff's lawyer. This is revolutionizing because as a plaintiff's lawyer, you're not being paid by the hour. You're being paid on a contingency fee normally. So if you can save time, that's great. But if you're an hourly billing lawyer like Bill, every, you know, 6/10, you know, every six minutes, this is not good for your practice. Even if you go into a hybrid type building or something, this may not be good if you're a flat fee. It may be because then, you know, you can save time, kind of like the contingency fees, but it's not always a good thing, especially for my hourly builders out there. It is not always a good thing because you've got to make sure you don't over bill the client. You only charge for the time spent and you just need to communicate. Communicate with the client. I cannot say that enough because otherwise if you don't communicate, we're never going to have consent. Rule 1.6 is going to be violated. 1.1, 1.4, 1.5. You're just going to violate all of them. So communication is going to be key. Now, how do you kind of help tailor this into the firm, create a firm policy, and we'll see this at the end of the slides, too. But just kind of the recommendations here, especially when it comes to fees and ChatGPT create a firm policy regarding ChatGPT its use or non-use in the firm always inform the clients if you're going to use it, an engagement letter or some disclosure, or have those conversations and outline how it's going to be billed and why. Maybe the billings can be different than they're used to seeing, especially if it's a reoccurring client. Communication is the key. We always need to communicate. Some other ethical rules that the last part of it, there's a few more that we could possibly violate with ChatGPT. The first one the next one actually is the one I call the kitchen sink rule. This is 8.4 and I call it the kitchen sink rules because it is so broad and it can be part of anything that it's easy to violate Section A if your jurisdiction has adopted it. We cannot knowingly assist others to do something we cannot do So if we cannot enter in client confidential information into ChatGPT, we can't tell our non-legal non attorney assistant to do it or a family member in our absence. We also cannot engage in the anything prejudicial to the administration of justice, lying to the court, not checking your sources, turning in false information, not communicating with the client. All these can be prejudicial to the administration of justice. See why I call it the kitchen sink rule? It kind of encompasses everything. And then we also to and this is another one where ChatGPT warns us about we cannot manifest by words or conduct any bias or prejudice based upon certain protected classes. This is kind of like the Title seven of the Rules of Professional Conduct, but it's broader than Title seven. But I will say not every jurisdiction has 8.4g. Some jurisdictions have rejected it altogether. Some have adopted it in full from the ABA model rule. Some have made it larger than the ABA model rule and some have made it smaller. So you really need to check your jurisdiction if you have 8.4g. But it's important to note this one, especially in regards to ChatGPT, is because we cannot discriminate while doing the practice of law for protected classes. Chatgpt tells us on its main screen that it may occasionally produce harmful instructions or biased content. If we just take what ChatGPT tells us at face value and turn it in, if you go back down to that rule 8.4 G, we could be violating that and having bias in our responses. So it is one to consider and some of the bias and stereotypes can come from just stuff on the internet. If it's pulling from any source on the Internet, then it could pull pull from bias content, from prejudicial content, from stereotypes, from anywhere at all. And it could be as innocent as only males or police officers, females or receptionists. Or it could be downright bigotry that it's pulling from. So reading what it produces you, making sure you understand it and making sure you don't violate the bias. Prejudice rule is also important. Now, the little trick to that is if you have the same bias as the response being generated, you're not going to catch it, which is always a danger because we all hold biases, conscious and unconscious, some more than others. But if you have a brain, you have a bias. And that's just how we're programmed. It's not a bad thing. It's not a good thing. It is just how we're programmed. So we need to watch out for it. The last rule is to kind of look out for supervision. I've already mentioned this briefly earlier in the seminar. But we do need to supervise our non-attorney assistants. The rules 5.1 and 5.3 really do require us to supervise, supervise and train other lawyers and non-lawyer staff. So if you are in charge of associates, you need to train the associates. If you have a secretary, a legal administrative assistant, paralegals, or in charge of the whole firm, everyone in it. You need to educate them on ChatGPT the risk, the benefits, and where we could violate the rules of professional conduct because they may think, Oh, it's this great new technology, we can use it and do with it what we want. When that is not the case. And so they need to understand the correlation between, you know, using the source and using it safely. And where we could really mess up in the rules of professional conduct. Because if they violate the rules and practice, you violate the rules and practice. So you really need to watch out for those things and just some things to consider. I've already posed this question earlier. Can you definitively say your assistant has not given you something generated in whole or in part by ChatGPT? Scary thought. Can you definitively say your experts or outside vendors have not given you something that was generated in whole or in part by ChatGPT? Scary thought. And do you even ask? Do you have that conversation with your vendors and experts and people you hire to help you with your case? Have you ever asked them do you utilize artificial intelligence or ChatGPT? It's a question you may want to add to your inquiries when you're hiring an expert or you're hiring a vendor, or you're hiring an outside third party to help you. It is something that you need to think about and consider because of the risk associated with their ethics. And if they violate it at our direction or without our direction, we just trust them. Then we violate it. So we really need to watch out for those that we associate with. Some other things just in general. Chatgpt. It can be abused and misused, just like anything else. It is starting to be used for harmful purposes and this has to go with more with data security of your firm and just cybersecurity. We are all know probably by now phishing emails and malicious actors out there giving fake news, trying to basically scam us into different things. Well, these cybercriminals are getting smarter or ChatGPT is making them smarter because they can actually use ChatGPT to help them in their cyber crimes. Cyber criminals can use the platform to produce content that appears legitimate since it produces it in that human conversational manner. If you think about it, when you got a phishing email, you could kind of spot it. The spacing was off. There may be misspellings. Capitalization was off. It was just a little wonky in that kind of stood out. Well, now cyber criminals can have ChatGPT generate those scam emails without all those mistakes in them. So it's going to start to be harder to catch things. They can also have ChatGPT write sales pitches for them, fake opportunities, fake job postings, fake you know, things to really scam you into turning over personal information or thinking you're dealing with a legitimate business and it's no more. You won't hear the poor grammar or the broken English or have the misspellings or any of those things. We're going to see cyber criminals be smarter and have better work product, which is terrible, because then we're not going to be able to catch it as easily. Some other things, too. And this is not just with ChatGPT, but just artificial intelligence in general. Cyber criminals are using it now to produce malicious codes that can infiltrate or hack other computers. It's actually ChatGPT version four, which is the one we're on right now that is more advanced and more sophisticated than any other version to date. Obviously, as they get more versions, they get more sophisticated and it can actually understand the context of coding and correct errors and coding, fix programming mistakes and develop codes for criminals or other people like cyber criminals. And so it's not always used for good. Like you may think, Oh, that's a great thing. It can help me in just coding in general and fixing computer mistakes, but in the wrong hands. Or when cyber criminals use it, they're going to use it for a harmful purpose. And that may be to come after law firms because we are one of the most hacked entities. Just because we house so much secure confidential data. We have clients names, Social Security numbers, our own employees. So we really have to watch out for those things. Some other things artificial intelligence can do that's really scary. Create fake news And they can also now there is artificial intelligence, not ChatGPT specifically that can engineer deep fakes, audios and videos, which means they can say, for instance, someone could take my voice programming into an artificial intelligence and then they can use my voice and the way it sounds to generate a message, send it to my loved ones. My loved ones think it's coming from me and it all be fake. Now, a lot of scammers use this to draw money or different things or fear or different things like that out of people. Get them to do something, manipulate them by cloning voices of loved ones, cloning videos, or making fake videos where you think someone's being harmed. And so artificial intelligence has these capabilities. And it's scary, really, to think about what our cybersecurity is going to look like in law firms and just in general as this technology gets more advanced, because as it does, it's going to be better at coding, it's going to be better at deepfake information and people using it for malicious purposes. It's going to be better at all of those things. And that is a terrifying thought when you really get down to thinking of the ins and outs of that. So what do we do? How do we mitigate? Obviously, firms need to put in place protocols. Procedures have checklists. Maybe it could be something as simple as if you're going to use it. You have to alert management. They need to sit down, have a conversation. You need to have a conversation with the client. Remember 1.4, keep them informed. Get everything consented to in writing risk and benefit. Sign off on it. You need to have some policy or maybe your policy or your firm is we're not using it. Not until it gets more advanced or not until there's a legal version, because there is not an 100% legal version out there. I have heard that there are different legal search engines like Westlaw and LexisNexis are trying to incorporate a ChatGPT type feature into them, but it's not on the market yet. There is no 100% legal confidential product just for lawyers, the ones that are out there more for general public. And we have all these risk and concerns for it. Now, will there be some in the future? Maybe. Hopefully we'll see. Provide feedback. Obviously have a reporting system training your staff if they see someone using it or not or misusing it, that's something you need to know. Talk with your vendors. Talk with the third parties you hire. Be extra careful and knowledgeable just about cyber security. Don't think you're going to see all those misspellings anymore. Don't click on those links in the emails. Watch out for those things and just continue really to educate yourself on ChatGPT what it can do, what it can't do, and how it can maybe benefit you and your practice or how it could really hurt you in your practice. And having that education and knowledge is vital to it as well. So these are just some things to think about. And there are different things. I mean, you could even if you like, I don't even know where to start on a policy procedure or protocol. You just start with the basics, like can you use it? If your answer is going to be yes, okay, under what circumstances? And just start going down from there and have lawyers do check points like why is it necessary? Can I get this from another source that's more reliable? Did I double check the accuracy? Like meaning if it gives you cases, did you check them against Westlaw? Just things like that really should go into the policy. I do want to at this point, let's try it out together before we get to the end of the seminar here. So we are going to hop over. This is ChatGPT. When you go to ChatGPT, you can type it in at the toolbar. You can see it at the top there on my screen. Chatgpt Openai.com. Openai is the platform that created it. You can see there, usually when you get to the first page, you sign in with your free account. If you don't have a free account, you used to have it, have it, email address and a password. I'd already signed into this. So this is the opening screen after you sign in. As you can see here on the opening screen after you sign in, it tells me things. It tells me examples of how to ask it questions. It tells me its capabilities. It tells me it will remember what you said in earlier conversation. So it is telling me it's acting like a black box and is logging everything I put into this. It can provide user follow up corrections and it can decline inappropriate responses. That is one thing it tries to do. You can't put in here something entirely inappropriate and you think it's going to come back to you and be funny. It will decline inappropriate questions and responses or it's supposed to, and then it tells you the limitations or some of the highlighted ones. First one on there may occasionally generate inaccurate information, Rule one, point one violation right there. We've got to know this is not 100% accurate. This is not our Westlaw or Lexis with cases verified where we know where they come from. This is pre trained information where we have no idea where it comes from or if it's true or accurate. It is telling us that there and also it gives us that warning. It can produce the harmful instructions or biased content, and that is limited to knowledge of events about that 2021 time frame. So all those limitations we've already talked about are warnings are on the screen and then it goes down and even says chat GPT down here at the very bottom may produce inaccurate information about people, places or facts. It tells you it twice on the screen. So we know it's not going to be 100% accurate all the time. I'm going to refresh it real quick and I'm just going to show you kind of a. So the first one I'm going to QE is say your assistant or you want it. Didn't know how to write interrogatories. You could say something, draft me interrogatories. For a personal injury case. Now, it's probably may come up with a disclaimer first, but again, please note these interrogatories are provided for informational purposes only, not tailored to your specific case. You need to consult an attorney. It does a disclaimer, but it still gives me an derogatory. As you can see, it's still going and when it's going to keep generating, keep generating them. It gave me 12 and we all know we can have more than that, just depending on sub parts. But if we look at what it gave me, state your full name and contact information. That's a general question. We do describe in detail the incident that led to the personal injury claim. It can draft it, just look at it. It's similar to the questions we ask as lawyers. It can draft it for you. And so lawyers are catching on to this. And I was actually sitting by one lawyer at a dinner once a few months ago, and this lawyer turned to me. We were talking about ChatGPT, and she says, Yeah, I use ChatGPT to draft me interrogatories because it can do it better than me. That took me under 30s. Now all I have to do is cut and paste, Put on the caption I've drafted interrogatories and less than six minutes. Maybe you can get a point two out of it. Think about your billing. I didn't even ask the attorney I was sitting with how she billed it or what she did. Or did she go in and tailor it or anything? She just said, Yeah, I use it to drop me interrogatories because they can do it better than me. Under 30s mind blowing and so many concerns with that. And I just entered a general question. Just think, if I would have entered a more specific 1.6 type of issue question where I gave a client name, where I gave client specifics, it would have generated out in grave detail. And that detail is the 1.6 violation. So we have to really watch out for those things. Another one we could say here. Let's do ten. Please give. Provide me. In case citations regarding unethical attorneys. We'll just put it that way. So let's see if it gives us so it's going to start generating, which we can't see it because I scrolled up. Um, so this one's going to say, you know, it can't do this. It's going to give legal opinions and stuff. Sometimes it does that. You've got to QE it just right. Otherwise it's going to give you a disclaimer like this and say it can't give me case citations. And so here it says, I'm sorry, but the internet or provide specific citations with my training only goes up to 2021. So depending on the question you ask and I will go to an earlier back from like May here. Or June or March. And so here's where I ask it for voir dire discrimination cases. And again, it gave me a couple, gave me Batson, which we know is real, and it gave me another one. And then I did another back in March saying, I want ethical attorney violations, know rule 1.6. And so it did say it does not have access to real time databases of case law to provide you with the most recent cases. But here are some that it's found in the past. Now, the thing is, when I take these cases and I've done this before, I have Westlaw pulled up. So I'll just start a new session in Westlaw here just because I want to show you that these cases are not real. Some of them are not real, some of them are real, but some of them are not real. So we'll start our new session here. We'll go back over to here and we'll just cut and paste. This first one, if I remember correctly, is actually a real case. But what you need to do each time, if you're going to use this is you need to check the accuracy of it so you can plug it in there. I'm just going to do all states, including federal. Um, just so we have very comprehensive search for these and it's going to look up and this one I think was real. I hit the search button. It's thinking and it says Aurora versus or United States versus Aurora. We go back over here. Okay, that's a correct case. We know that one's real. Now, do we know if it's real? Based on the description here, we'd have to read the case and see which we don't have time. Then you need to go down to the next one. You shouldn't just say, Oh, the first one is real. All of them are real. So we're going to cut and paste the next citation in there, see what comes up. This one's not real. That citation doesn't even exist. This citation was 210P3D Pacific third 1069, 1069 is actually, it looks like the middle of a case state versus Jones out of Washington. 210 Pacific 1068 So that is not a real case because that is Jones State versus Jones. And this is telling me it should be in Ray Miller from Oregon of oh nine, and this is from Washington in 2009. So the second case is not real. It's a fake case. And so you can see it doesn't always give us accurate information and we can even keep going. And this is what that lawyer in New York that got sanctioned failed to do. They failed to double check their cases. They failed to double check the accuracy of the information. Now, if I was going to be using this information, here's the next one. It's a fake case as well. And if I was going to be using this information that ChatGPT is giving me for a brief, I would need to go out and do this for each of them, read the cases like print them off or read them online, and then make sure even the summary is accurate because even the summary could be wrong. This gives you inaccurate information. I cannot stress that enough. It's going to draw violations right and left. 1.1, 1.6, 8.4. All of them. You need to be very careful with this. Is there any use for lawyers to use this? The answer to that is yes. Lawyers can use ChatGPT for things, but I would not use it to do your legal research and think of it as gold or the money on that. I would not use it to draft your motions in that because it's going to be an accurate and very general. What I would maybe recommend using it for something that would be safe is more like a starting point. How many of you have ever gone into Google and just want something to bounce off of or, you know, you don't know where to start or you don't know. You even know how to phrase the question. So you just enter a random stuff in Google and see what comes up. You can use this search engine for legal stuff kind of like that, just as that starting point. But it should not be everything. Doing it all for you, doing it all for the clients type of thing. That is not what it was made for. It is not a legal search engine and we need to always be careful. But it does draft things like if you need a poem, it can draft you a poem. We probably all wanted that, you know, when we were in grade school. But just something to really consider. We are kind of winding up on the hour. If you do have any questions or anything, I'll put my information back up here on the screen. Please feel free to contact me at any time. You can also reach out to the platform. They know how to get in contact with me. Happy to answer any questions. Happy to speak with you more about ChatGPT or artificial intelligence in general. And thank you for listening here today and I hope you have a great day.