“AI, Equity, and Access to Justice: Practical Guidance for SRLs and Legal Professionals”

Video recording can be accessed here

Recorded March 13, 2026

*PLEASE NOTE: This is a rough transcript of the live webinar event, automated by OtterAI. The transcript has not been edited or checked for accuracy against the recording.

Speakers:

Transcript:

00:14

All right, welcome everybody. I’m going to get started right on the.of the hour here, because we’ve got a lot to cover and we’ve got a kind of a tight agenda. But first of all, I wanted to say welcome to all of our attendees. We’re really grateful that you are joining us today for AI equity and access to justice, practical guidance for SRLs and legal professionals using generative AI. My name is Dana Cornwall. I’m project manager at the National self represented litigants project, and we’re really excited that we’re partnering today with the Association of Black paralegals.

00:53

Before I move on to introducing our moderator, I just want to take a second to

01:01

do a brief land acknowledgement. So we acknowledge that we are gathering on the traditional and ancestral territory of many nations, and that the land we now call Ontario is home to many diverse indigenous First Nations Inuit and Metis people. Our presenters today are speaking from the Toronto area, which is located on the traditional territory of many nations, including the Mississaugas of the credit the Anishinaabeg, the Chippewa, the Haudenosaunee and the Wendat peoples. And they’re speaking from Waterloo, which is of the Haldimand tract, within the territory of the neutral Anishinaabe and Haudenosaunee people. I myself am located in Windsor, which is on the traditional territory of the three fires, confederacy of First Nations, which includes the Ojibwa, the Ottawa and the Potawatomi. We acknowledge and celebrate that Ontario and our communities continue to be home to many diverse First Nations Inuit and Metis peoples, as well as people from many other cultures and ethnic groups, wherever you may be joining from today, we encourage you to reflect on the history of the land you are on and the ongoing impacts of colonialism in your community. In the context of today’s session, in particular, we would also like to acknowledge the historical and ongoing impacts of colonial legal systems on indigenous peoples throughout Canada and on the communities who have experienced marginalization within those systems.

02:27

So I’d like to introduce our moderator for today, Liz conliff,

02:35

sorry, Liz, I pull up your bio here, too many things open on my screen. So Liz is the owner and founder of lischen paralegal services. She’s a licensed paralegal with over 20 years in the legal field. She’s also a member of the Law Society of Ontario and the Ontario paralegal network, and she is the chair of the Ontario Association of Black paralegals. She holds designations as a licensed paralegal family law service provider and human resource manager. In addition to her legal work, Liz contributes her time to the program advisory committee at the Academy of learning, helping to shape the next generation of paralegal professionals and to level justice, assisting in educating the youth of our future about criminal justice. So I’m going to turn things over to Liz, who will introduce our panelists, and then we’ll get going with our great content today. Liz,

03:24

thank you so much. Dana, thank you for inviting me to this wonderful webinar about AI equity and access to justice. Now this webinar is allows us to talk about the reason why it is so important to use AI when we think about legal services and court related processes. And this session is designed for self rep individuals as well as legal professionals who want to clear, really a practical introduction to AI and what it can and cannot do and how it can be used responsible, especially when preparing for legal matters. Now, today’s presentation is general information only, and it is not legal advice. Now, AI tools can definitely make mistakes or miss important contacts, so you should always verify information using reliable sources and where possible, seek legal assistance. Please do not enter confidential information when you are using AI tools, unless you understand how your information is being stored, used or shared. Now, in the next little while, while we are online here for this webinar, we’ll cover just some AI terms in plain language, some examples and helpful tips on how to use and definitely common risks and definitely some tips for drafting and documents and so on and so forth. So questions can be asked, and we ask that you put them in the chat so that we can definitely answer any questions that you may have today. Now there will be time for Q and A at the end, hopefully we don’t run out of.

05:00

Time today, this

05:03

webinar is being recorded for those, everyone that’s watching to be aware, so that if you want to watch it at a later time, you do have that opportunity. I want to thank you all for being here, and now I’m going to take the time to introduce our wonderful panelists. We start with Mara R Grossman, and she is a research professor in computer science at the University of Waterloo. She’s also an adjunct professor at Osgoode Hall Law School and an affiliate faculty member at the vector Institute for artificial intelligence. She is widely recognized for her research on the use of AI and machine learning and legal contacts, including technology assistant review and responsible AI

05:45

Mara also serves as a legal expert in high profile court cases, and is the principal of Mara Grossman law. We then have Annette Demers. Annette is a law librarian and a legal researcher with decades of experience in legal information and access to justice. She is the former head of the Paul Martin Law Library and currently chairs the Canadian Association of law libraries in the artificial intelligence Working Group. Annette is also a past president of call and a leader in legal research education across Canada. We then have Lisandra conleaf. Lisandra is an educational developer at the Ontario Tech University and a doctoral student at the University of Calgary. Her research focuses on ethical and inclusive use of AI technology, particularly how AI can support equity and access without reinforcing systemic biases. She brings extensive experience in education, digital technology and equity focused leadership. So that being said, and now with the introduction, we are going to start with our wonderful session today, and we’re going to start by diving right into some questions. We’re going to start by setting the frame, and it’s aI beyond the hallucination panic. So that’s going to start with Lisandra. If I can get Lisandra to join the conversation.

07:13

Thank you so much. Liz, thank you. Now as Lisandra, we hear a lot about hallucinations. Now, what do you think the real risk is when the courts look at AI use?

07:25

So I think when we talk about hallucination, I know Mara is going to be going into this in more detail in her presentation, but I wanted to focus more on when you’re talking about hallucination with the courts, it’s really looking at the information that you receive and how you’re going to vet that information, because not all the time,

07:52

what is produced by AI, or what is generated by AI is factual, and I know that Maura will be going into that more deeply when she’s talking about her presentation.

08:04

Thank you. Alessandra,

08:06

now is there an expectation that we have for uses about the AI tools today?

08:14

There is,

08:16

I’m thinking it might be better if I start off with my presentation. Sure How I can actually balance AI with human professional values, and that would kind of take us into any additional questions that we have. Excellent. Thank you for that. All right,

08:38

so we’re going to look, basically today, at balancing AI with human professional values. And we’re going to look at the benefit and the risk where there’s some EDI consideration, professional standards. And then at the end, Annette and I will jump back into SLRs and some advice that we can give.

08:58

So setting the stage. So here where I’m going to look today at what AI is doing in the legal practice, the risk, look at some of the benefits, the EDI lens, professional standards, SLR and AI. I’m also going to look at just a pathway forward.

09:16

So we know that AI adoption in the legal profession is no longer speculative. It is already embedded in everyday practices and in simple, trusted tools that you already know. So through a recent industry survey, it’s indicated that about 74% to 80% of legal professionals report using AI, and about 69% are using it for a general purpose, a system in their work or they show it may show up in their shift from experimental or piloting use. Sometimes it’s also being integrated into routine workflow. For example, if you’re using a Google workspace or

09:59

maybe.

10:00

Be a Microsoft office workspace, AI is already embedded in those tools.

10:05

So while adopt adoption, the data also reveals that there’s a significant governance gap. While adoption is what widespread and only about 9% of law firms currently have a form formal AI policy or governance framework in place, in contrast to in house legal departments, which appear further ahead, with approximately 96%

10:29

allowing for some form of AI to be used, compared to traditional

10:34

law firm of about 74%

10:37

so what does all this mean?

10:42

So the these tools can enhance the legal profession, and I stress on the word enhance, because all technology tools that you are using are only created to enhance the human ability to complete your task. You are the professional, and you’re still responsible for the validity of the final information put forward or presented. So I will say, for AI tools used primarily to argument your routine, efficiency, efficient, efficiency of time intensive legal tasks. But it should never replace your professional judgment.

11:18

The most common use, according to current research, includes League of research, and we talked about that 74%

11:27

where they help identify relevant precedents and synthesize case law, then you have the document review and drafting and about for summarization, About 77% usage when dealing with large volumes of material, and then the access to justice.

11:47

AI tools can help lower income lower income clients understand legal documents, and they can also expand their legal capacity there for reducing barriers

12:00

on the administrative relief side, these tools do translate into measurable productivity gains. The report also suggested that AI tools can save lawyers or professionals approximately 240 hours per year, and other professionals are also reported that it also improved efficiency. But the problem here is that as much as these tools can give us research, efficiency assistance in documenting, reviewing and drafting or supporting lower income with understanding legal terms, we also have to be mindful that everything needs to be vetted over and over, and I know that Annette and Mara would go into that in detail.

12:46

So what you really must guard against looking at the risk is, as a legal profession, you must guard against these interrelated concern when using AI and four major concerns when using AI, one is accuracy and the risk of hallucination. And when we talk about hallucination, it’s information that is poured out, and as I said, Mario will jump into that more information that is generated that is not factual. We also have to look at confidentiality and privacy. So uploading

13:22

client data into public. AI tools may breach some privilege, so you have to be very careful what you’re uploading. So the central shift is not merely that AI makes mistake, it is that it can generate errors with high confidence and present them in forms that resemble legitimate legal authority, and that’s where the accuracy and hallucination comes in. So this creates a heightened professional task for lawyers and legal professionals.

13:51

You might be aware that recent Ontario practices direction effectively place AI usage within a professional conduct framework, meaning lawyers and other legal professionals remain fully responsible for the accuracy and submission and you must independently verify all authorities a defensible workflow. Therefore treats AI outputs as preliminary draft rather than reliable sources. So every citation, every quotation and legal proposition, you must check against primary law or trusted legal databases. So there has been a few Canadian cases that illustrate these consequences. And regardless of whether an error originates from Ai counsel or legal profession, retains the duty not to mislead the court. And I know Maura is going to give you some examples of that

14:42

for confidentiality and privacy, the risk with AI extends beyond just simply naming a client. Even when identifying details are removed, factual patterns can still reveal a client or the matter, and because many AI tools retain these prompts or use them for.

15:00

Model improvement, uploading documents may inadvertently disclose confidential information. So you should treat AI vendors or like third party services providers review Terms of Service, retention policies and training practices so where confidentiality can be assured inform clients, consent or careful reduction may be required. So you should always limit the use to approve tools, to only tools that are approved, and therefore it prevents you from encountering what we call a shadow of AI

15:36

on the ethical and competence obligation professional competence in Ontario increasingly includes technology competence.

15:44

So all legal professions must understand both the capabilities and the limitations of AI tools and how the use of them therefore intersect with duties such as confidentiality and supervision. Generative AI cannot replace professional judgment. Its output must always be independently verified.

16:06

The thing that we face a lot with AI that we don’t talk about is the systemic bias, and that’s where my research come into play. And this is why I was very curious to see what how AI tools are being

16:19

how their data is being sourced, and this kind of lead me to where I am doing my doctoral research. There are many systematic biases in central ethical concern with generative AI. Because these tools are trained on large historical data set, they can reproduce and amplify existing social and legal biases. So basically, bias may appear, not only in the automated decision, but also in the summarization, the drafting or the issue framing. So I would therefore advise you that seek guidance from the Canadian legal

16:54

recommendation and how they talked about treating bias as a quality control issue, always testing the outputs across scenarios, incorporating human review at the end of it, we always have to be the human in the loop where we have to review whatever documentation we’re presenting.

17:15

My sort of my passion piece is on the equity and diversity and inclusion of AI assisted law. So AI tools become more common in legal research, drafting, case management, as mentioned before, but we need to consider the equity, diversity and inclusion implication. AI can improve efficiency and legal work, but we need to ensure that it does not reinforce systematic inequities. So I wanted to outline some of the key EDI risks associated with AI used in legal practice. One of the biggest risks AI tools perpetuate is algorithmic bias, and AI system learn from historical data where this may reflect past inequities, and it means that AI can unintentional, reproduce discriminating patterns. For example, some risk assessment tools using criminal justice have demonstrated racial biases. So those are some of the things that may be replicated. There’s also the risk of language and culture barriers. AI tools tend not to perform less accurate with non standard English or diverse linguistic patterns, and it may miss interpret important culture or societal context. Another risk is access to justice concerns. So with increased reliance on AI tools, it actually widens the digital divide, and therefore disadvantage the client with limited technology access to digital literacy.

18:48

Finally, AI can introduce workplace equity risks such as gender language patterns that may disadvantage women in hiring or evaluation systems, and the confidentiality risk is if sensitive client information is entered into public AI platforms, so transition to best practice. So how do we address these risks? We need to maintain human oversight and critically review all AI outputs for bias or inaccuracies. It’s also important to audit AI tools for biases and request transparency from designers and creators about training data and limitation at all times, these AI tools should supplement, not replace, culturally competent legal counsel and professionals must protect client confidentiality when using these tools, we need to also develop internal policies and training on responsible AI use, and therefore a centering marginalized client perspective in AI governance is also very important. We also need to advocate for diverse developmental teams and includes the data set. Basically AI can also be used positively.

20:00

To expand access to justice, for example, through translation tools or plain language legal summaries.

20:09

So

20:12

how do we preserve those judgment

20:15

for ethics and human expertise? I say one, we have to look at the competency Rule three, dash one, dash two. Of the LSO rules requires ongoing competency, and it includes understanding the tools you use. So it states professions must understand AI’s limitation, not just its output for supervision. AI is not a junior associate. It is not also your assistant. It has no judgment. All AI generated work must be reviewed, and I think that’s the common thread here, verified and signed off by a licensed professional who takes the responsibility. And when it comes to confidentiality, clients data must be protected, especially under Section 3.1 before using any AI tool, confirm where the data is stored, who can access it, and Does the vendor have adequate protection? And I just wanted to say a I should make lawyers more human, or legal professions more human, not less. Your judgment, empathy and ethical compassion remain irreplaceable.

21:16

Thank you.

21:21

Thank you very much, Lisandra for that key insight. I heard so many things screaming at me, access to justice, concerns responsible, AI, use and benefits versus risks. Thank you. Thank you. Thank you. I’m going to move on now to Maura so that she can go into a little bit more detail about accuracy, reliability and some of the limitations that AI has with our industry.

21:50

Hey everybody, thank you. Liz, a delighted to be here. I’m going to talk a little bit about hallucinations. I’m not going to use any slides today. I’m not as ambitious as my two co presenters. Actually, I’ve been traveling a lot, so I wasn’t able to do any slides, but I don’t think we’ll need them for this topic. So why do large language models hallucinate? Well, we we read them, and we think, you know, we read the responses to prompts. And we think these tools are thinking, but they’re not thinking. They are. They’ve been trained on the internet, and they’ve been sort of maximized to optimize fluency. So they’re they’re very, very good at coming up with what’s the next most likely word to fill in this blank here, because they’ve been trained on the entire internet and large proprietary databases of books and magazines and government writings. So they’re very, very, very fluent. But unless they’re they’re set up a particular way which most of the public facing tools are not. They don’t really have any way of checking what they’re saying. So if they’ve learned, if half the information on the internet says cinnamon causes can cures cancer, and half the information on the internet says you need chemotherapy or radiotherapy or something like that, they may not know which to answer, so they may Guess. They’re also trained to give you an answer, rather than to say, I don’t know, or I haven’t seen the answer to this, or I’m not sure, they’re what we call sycophantic. So they’re, they’re aiming to please you and to give you an answer. So if you ask for something it has seen many, many, many, many times, like the US Supreme Court opinion in Roe v Wade

24:11

that had originally permitted abortions. It’s probably read so much about that case, it’s probably going to get it right. But if you say, I want an Ontario nuisance case where, you know, one neighbor gets evicted because they’re making too much noise, or they’re cooking food that smells horrible, it’s not going to have seen that. It’s unlikely to have been trained on that it wants to give you an answer, so it’s going to make it up. So we have to understand that’s the nature of the beast. It’s it’s very different when we’re dealing with commercial tools that have been trained on on Database.

25:00

Basis of case law. These are different because, unlike the tools that many self represented people are using, or if you just go to chat, GPT, anthropic, Claude, you know, Gemini from Google, those are not legally trained. They’re not specifically tuned on legal cases, but if you get one of the commercial tools, well, what it does first is it does the search to find the answer to your question, and then it goes into the database and it sort of checks is that the right answer? Can I actually find this case? And then often, it’ll give you an answer with an actual link to the case. Now, the commercial tools still make errors. They’re just different kinds of errors. So whereas chat GPT might just completely make up a case for my my nuisance case, Westlaw and Lexus aren’t going to do that. They will actually find a case, but the kinds of errors they can make, maybe it’s not an Ontario case, maybe it’s a BC case, or maybe they’re quoting from the dissent, not the major opinion. Maybe they’re mistaking what the trial court says for the appellate court. So even the professional tools hallucinate, they hallucinate less, and the nature of the hallucinations are different. So what does this mean for you? As Lisandra said, check, check, check, check, verify, verify, verify. So if you I wouldn’t recommend using chat, GPT or Claude or or something like, you know, open and free for doing legal research, but if you are self represented, you may have no choice. So

27:00

if you do do that, you you have to take that second step of going to Canley or going to Google and finding the actual case. And not only just finding it and saying there’s such a case, it didn’t make the case up, but you actually have to read that case and make sure it says what you’re saying it says, because these tools can also misrepresent a case, and that’s just as bad or they can. They can screw up quotes, so we always have to site check and not just the actual that the case exists and then it’s got the right state citation, but we have to check quotes, we have to check facts and look for any other kinds of errors. Now, if you are self represented,

27:59

you may get a little more slack from the court

28:06

if they think you are operating in good faith. You’re writing your own brief, and you cite a case that doesn’t exist. You may or may not. Me as a lawyer, I’m going to get in big trouble no matter what. Or the paraprofessionals, because they know we have access to, can lead to Westlaw, to Lexis, and we can check these things now. Self represented litigants do have access to, can lead for free, so you should be checking there, but most courts have cut a little slack, a little more slack, not much, and I wouldn’t count on it. I would assume that I could get punished. And the punishments for lawyers and legal professionals is you can be reported to a disciplinary agency that governs lawyers or paraprofessionals. Your client’s case can be dismissed.

29:13

You can lose credit, complete credibility with the court. There can be fines and all kinds of things for self represented litigants. When the courts get angry enough, they will impose fines.

29:29

They may impose costs. In other words, they may make you pay the other side’s legal bills for having to find out that the case was fake and that you cited it and reporting it,

29:44

or they may dismiss your case. They may just decide, I’m not going to bother with this. I you know, if the person can’t even cite the cases correctly or state the facts correctly, so you really, really, you know, have to be.

30:00

Be cautious and careful in this area. So what can you do? Well, ideally, if you have access to

30:11

something that is trained on legal cases, that’s the best thing to use, always the best thing to use. It may still make mistakes. It’s not going to make cases up out of whole cloth. If you’re on a general purpose tool, that’s a much, much bigger risk. But like I said, you may not have access now in Canada, can leave is free, so there’s no reason you shouldn’t go to Canley and check every single case that you are citing.

30:45

I think that’s pretty much my comments on

30:50

hallucinations. Liz, unless you have any specific questions for me, or somebody in the audience has a specific question, I can certainly go into more detail.

31:03

I do have one question for you, Maura, when we think about hallucinations in court and why they are so dangerous,

31:11

is there like a problematic example that you might be able to give in a legal setting? Sure? So this has happened more than once.

31:25

The parties prepare an order for the court to enter, and it cited a case, except that case doesn’t exist. Now the court enters the order with the fake case, and that becomes law, except it’s not law because it’s not based on a real case. And then another judge comes along and copies that or quotes the the new case. And so there is a risk of really perverting, perverting the law of people’s losing confidence in in the law. But there have been three or four examples where courts either on their own or taking materials from

32:17

things that were cited by the parties and and citing them in opinions, and that really, really harms the law, because now we’re relying on things it for precedent that that don’t even really exist. So that’s one of it, and also people’s confidence in the judicial system, that it’s going to be fair and accurate and that we can count on, on what lawyers or self represented litigants say to the court, but, but courts will have to go much, much slower if they have to actually check every single thing you’ve said. I mean, we’re already backlogged, but imagine how much more backlog we’d get if, if the court has that much more homework to do thank

33:02

you for those wonderful examples. Myra, I’ve got one other question for you there potentially. Or could you point out maybe some red flags that users can watch for when reviewing their AI outputs?

33:16

So

33:18

this one, this one’s hard.

33:23

But if you were asking for something somewhat obscure and you get the magic, perfect answer,

33:33

I would be a little skeptical of that, like you’re saying I need I live in Brampton, and I want a case in Brampton that says this, this, this, and this from 2024,

33:47

and up pops, exactly what you’ve asked for. I might be a little skeptical of that.

33:57

Really, it’s the verification that’s the most important thing. And the problem is, and I didn’t mention this before, but it’s not totally a hallucination problem, but it’s easy for me to check

34:13

if I’m asking the system,

34:16

write a chronology for me of all the events that happened, and then link back to a document to show me that event really happened, or that Liz really said this. Link back to her transcript, whatever. So I can click on Liz’s transcript, and I can see Liz actually said that at page 37 line 42

34:39

but sometimes when I ask for a chronology, it leaves something out,

34:47

then I don’t have anything to click on, and something really important can be left out. So that’s a really tricky one that’s not so easy to find. It’s easier to click, click.

35:00

And I would encourage people, if you are going to use the general purpose tools like chat GPT, ask the tool to provide you a link or a copy of the opinion or a link to the site so that you can actually click on it and check but the place where you can get yourself in big trouble is where there’s nothing to click on and you don’t know that you missed something. So you’ve asked

35:29

take this case and tell me all the key claims. Now you’ve asked the tool to exercise judgment and choose which are the key claims from which are all the claims? Now it’s not thinking like you or I think so. It doesn’t know what’s important and what’s not important. It might ask itself,

35:53

which was given more space in the papers? It’s given more space. It must be key. Which one has a title. It’s a title. It must be key, and that’s not how lawyers think. There can be a very obscure fact that happens that changes the whole nature of the case, but it isn’t given a lot of space, or it doesn’t isn’t underlined in the title. So when you ask the system to exercise judgment in some way, that’s very risky, so safer tasks, summarize all the claims,

36:33

or tell me which claims the court dismissed, or what was the court’s argument for why they decided. This case this way, you tend to get better answers when you ask these open ended questions or ask the tool to exercise some kind of judgment that’s much riskier.

36:55

Thank you, Mara. And so what I am loving this so much, and I hear read the case, use reliable search engines and cite check. When we think about governance, we’re thinking about the ethical responsibilities that we have as licensed professionals, and so the leniency that they may give to self reps is not given to the same as legal professionals. And I want to thank you for pointing that out, and now we’re going to take an opportunity to introduce Annette. Annette, I’m going to bring you on and ask you about AI uses in legal practice. I guess the model selection and some legal specifics on free users, very good. Thank you so much. I’m very pleased to be presenting alongside my knowledgeable colleagues. It’s been fantastic so far, and thank you for the invitation. I’ve been watching AI’s use by self represented litigants for some time now, and my interest is in making sure that we all have stable and transparent access to legal information. So to help me stay current, I receive a regular email from Canley that alerts me anytime there’s a new judicial decision where litigants have been warned about submitting hallucinated cases to the courts and anywhere a new AI policy has been released,

38:20

just moving my slide forward here,

38:26

trying to use Zoom’s functionality here, there we go. Okay, perfect. Let me just go back and just had a glitch on it.

38:34

Okay,

38:36

so I have a whole mixed bag of things to tell you about today, and I hope there will be some interesting takeaways for everyone. My portion of the talk is mostly aimed at self represented litigants on the line, but I do have one slide that paralegals can take note of, and it’s a bit of a nod to some of the things that Maura was saying. So first I’m going to go over some basic definitions, so that we’re all generally on the same page, and then I’ll mention some facts and figures pertaining to AI’s use by self represented litigants in Canada. And finally, I’ll spend some time talking about the strengths and weaknesses of using AI for very specific purposes, as well as some things to consider before using it. So what I’m going to do now is just take a couple of minutes to make sure that we’re all on the same page as we move forward by giving you some definitions. So this next slide is for the self represented litigants who are on the line.

39:30

So in any case before the courts, there are two main types of questions that the judge is trying to answer. We call these questions of fact and questions of law. So questions of fact are the questions that the judge has to decide to determine what actually happened. Sometimes the parties might agree on what happened, but if they don’t agree, then the judge has to decide what happened. So how does a judge decide what happened? They.

40:00

Use evidence, such as documents, witness testimony, photographs, recordings, etc. The parties submit evidence to the judge, and it’s up to the judge to decide whether to admit the evidence and how much weight to give to it. Not all evidence is admissible. However, judges use the rules of evidence, such as the Evidence Act to decide what evidence is admissible. Now looking down the right hand side of this slide, questions of law are the questions the judge has to answer about what law applies. Sometimes this might be straightforward. So for example, the question was she required to stop at the stop sign would be answered by looking in Ontario at the Highway Traffic Act Section 136

40:46

However, for many areas of law, it might be more complicated to decide what law applies. So to decide what law applies, the judge will look at legal authorities such as legislation, regulations and judicial decisions or case law, to name a few. But the parties have to submit these legal authorities to the judge and try to persuade the judge about how to interpret the law in their favor. All legal authorities must be verifiable, so ultimately, why I’m telling you this is that authorities, and this is everything. Maura was just saying, authorities generated by AI that are not verifiable are called hallucinations. So whenever all three of us are referring to hallucinated cases or hallucinated authorities, we really aren’t meaning any of the following. So it could be an authority that does not exist at all. So AI made it up, which happens all the time. More explained how hallucinations can also come from relevant authorities that do exist, but the paragraph that you’re pointing to actually doesn’t support your argument. Or we can also be referring to authorities that do exist, but which are completely irrelevant to the case. So now that we’re all on the same page, it’s important to recognize that judges will not consider evidence that was generated by AI, nor will judges consider legal authorities generated by AI. The judge will disregard them and may penalize you, as Maura said, for submitting them. So what can you use AI for then in your preparations as a self represented litigant, I’ll talk about this in a little bit.

42:26

So as I mentioned, I’ve been tracking cases where lawyers and self represented litigants are being warned by the courts for submitting hallucinated cases.

42:35

Just do this. Okay? So the figures that you’re about to see are my tallies of judicial decisions and tribunal decisions from across Canada where litigants have been warned. All of these are listed on my guide to AI regulation in Canada, which is available through the University of Windsor Law Library website.

42:55

So according to my research, as of yesterday, in Canada, we now have 76 decisions where self represented litigants were warned for submitting hallucinated cases. This includes 10 cases where costs were assessed against the offending litigant. Generally, the costs are being awarded at about $500 each, with the highest assessment being $5,000 in a Quebec case, submitting AI generated evidence or legal authorities to the court is considered by many judges to be an abusive process, meaning the offending party who does this is wasting everyone else’s time, and the courts will not tolerate it.

43:35

So this slide is for the paralegals to take note of this decision was released yesterday. Kapahi Real Estate through the Ontario Superior Court, Supreme Court, excuse me, in this case, the lawyer of for the respondent delivered a factum that cited real cases with the correct neutral citations to Canley, but then Council added seven quotations that actually do not exist in the cases, the quotations were fake. So the lawyer also did not certify the authorities in his factum, as required by rule 4.0, 6.1,

44:11

subsection 2.1, of the rules of civil procedure. So he did not certify the authorities. And then the lawyer proceeded to deny using AI in the preparation of the factum. So the judge was either had to decide whether he lied about not using AI or fabricated direct quotations, both of which would be egregious. So ultimately, the judge indicates that he will be referring the matter to the Law Society of Ontario. And the judge also mentioned that in a prior case that he heard ko V Lee, which is pretty famous now among some of us, he the judge says, I ultimately referred prosecution of a contempt of court proceeding to the attorney general. So the stakes are now getting very high here for legally regulated professionals. And just to say, we all need to.

45:00

Sit up, take note and buckle down. This technology, when not used properly, can destroy careers.

45:08

So in the cases I reviewed, if you submitted hallucinated materials to the court, and Maura already mentioned this, some of these things so the court might strike out or disregard your submissions or portions of them, and this is the tool that they’re using the most often against self represented litigants. From what I’ve seen, they might order you to declare whether AI was used. They might direct you to provide copies of cases that you refer to or to provide a written explanation about why you used non existent case law. They might direct you to amend or redo your submissions. They might direct you not to use AI and further submissions, or ultimately, as Maura said, they could order you to pay costs to the other side or to the court for wasting people’s time, because ultimately, someone has to verify what you submitted. It’s your responsibility to verify what you submitted. It’s not anyone else’s responsibility.

46:06

So remember, I said that you can’t use AI to generate evidence. So here are two examples where litigants submitted evidence that was generated by artificial intelligence. So this one is an example that was reported in the decision of the BC civil resolution tribunal, and the issue is raised at paragraph 11, which says, Mr. I am in provided an overview of the Insurance Corporation of British Columbia’s policies prepared by generative artificial intelligence. So ultimately, the court disregarded this submission. They are not going to take a summary of a document that actually exists. And I’ll give you some more help on this sort of thing, navigating this in a future slide, another example from the Ontario Animal Care Review Board, again, citing paragraph 13. So I’m telling you now to if you wanted to read this case, you could take this actual citation and put it in. Can Lead and go read paragraph 13. So when you’re presenting materials to the judge, you can also say, please go see paragraph 13. This is how we indicate to the judge how we want them to read the authorities, or where we would like them to look in the authorities that we’re giving. So at paragraph 13, Mr. Lapointe’s use of artificial intelligence tools has produced false or misleading information. This information includes veterinary opinions that are likely fabricated and citations to non existent cases or cases that do not stand for the legal principles for which he relied on them. So this is another example of evidence that the judge will not so ultimately, the the court indicated that the rules of evidence would render AI generated evidence inadmissible, meaning that the court will not accept it. And ultimately, the Court also said that it could not use the fake legal authorities.

47:57

So this slide just highlights that you cannot and should not use AI to generate evidence. So this table shows you, on the left what evidence you should not try to generate, and on the right what you should be submitting instead. So everything on the right hand side is my generalization. So take it with a grain of salt. It’s likely more complicated than what my table says, and it’s the rules of evidence that govern all of this. So it’s just a generalization on the right hand side.

48:26

Ultimately, you should not be submitting AI summaries of documents, contracts or policies. Instead, you would submit the original you should not submit AI generated reports or opinions, especially an expert report. Or pretending to be an expert report, you would submit an original report from an expert, an AI generated image or a top touch up of an image. You should be submitting the original photos or images, AI generated videos or audio files or voicemail recordings, etc, those have to be submitted in an original file. As I say, generalization, there are rules of evidence governing all of this.

49:11

You should not be submitting AI generated or summarized emails or other correspondence. Provide the originals. AI generated business records, such as invoices, logs or financial statements, you should be submitting the originals AI generated social media posts. This is even more complicated. I’ve actually given you a link on my slide. I can make the slides available to go and read some more specific information about the rules of evidence with respect to social media.

49:41

One last slide on this vein is that we are also seeing decisions where judges are commenting on the appropriateness of using AI for certain purposes. This decision came from the BC workers compensation Appeal Tribunal where the employer was asking the court to disregard the self represented workers materials because he admitted.

50:00

That he had used AI in preparing them. But ultimately, the judge said,

50:06

I note that much of the workers written submissions mirror the discussions he had with the board, and I find it’s not the case that the substantive content was generated by AI. Instead, he largely used AI to organize his evidence and organize his submissions in writing. And so the judge did not listen to the employer and gave some weight to what the employee had submitted. So this is an example of appropriate uses. So what I did was I put together a little summary here

50:40

for writing purposes. If you’re just using chat, GBT, copilot or Gemini, you can use it to organize your thoughts and arguments into a logical order, to summarize your situation briefly in a logical order, to correct your spelling and grammar, to check change the tone of a sentence or a paragraph, or to polish the language that you’re using on the research side, maybe if you go to can lead, you could

51:09

get a summary for yourself of what a case has to say if you’re taking that summary from can lead. And just remember that can lead also has an AI interface. Now it’s called can Lee, Search Plus, it appears on the right hand side of can Lee, and you do need a free registration for my can Lee, so just log in for free.

51:32

This. These are products as Maura mentioned, these. This is legal authorities. Can Lee is a database of legal authorities, so you can try to see what can leap Search Plus can do for you. But ultimately, on the left side of the of this slide, I remind you that if you are going to use AI to help with drafting, always clearly state this at the start of the materials, including what AI you used and what you used it for. That way, the judge is going to know that you used it, and they’ll know how you used it and that you’re being transparent about it, and that will know that you’re not let them know that you’re not trying to deceive them.

52:09

I also have slow slides about what’s inappropriate uses, especially when it comes to legal research, so I would say it’s not the best idea to try to use chat GPT to figure out the correct court form to use. Instead, go directly to the Court’s website and see if you can find it that way. If you are trying to work with legislation, statutes, regulations, use can lead, as Maura said, do not use chat GPT. There is, there’s not a lot of Canadian legal authorities available for free on the web, and that’s what chat GBT and Gemini are using. Are the free sources. And you might say, well, aren’t they using Canley? They are not permitted to scrape canly. So chat GBT, Gemini and copilot are not permitted to scrape Canley. They are using anything else on the open web, which is usually us case law and US statutes, so just be super careful about that. Get your legal authorities from Canley, as Maura said,

53:11

also, I would not use chat GBT to scan legislation for relevant language. Instead, go into Canley review the statute. You can use the find function, but go to a library, go to a law library and ask them for some help. They’ll give you some books to read and that sort of thing.

53:31

For locating case law, you can try to use Canley Search Plus.

53:36

And then for preparing citations, you could ask a law librarian to help you kind of review your citations, and ultimately, if you’re trying to tell the judge where you’re getting your information from, which we call a pinpoint citation, so judge, please see paragraph 13. Make sure that you’re only citing those paragraphs if you’re sure the paragraph is number is relevant to what you are trying to argue.

54:05

And I’m going to call this my last slide, but I do have other material I can provide at the end if we run out of time from the rest of the presentation. But again, on legal research, do not use chat GPT to summarize legal issues or to summarize elements of the case that’s going to be proved.

54:24

I would say again, go to a library, look through textbooks and try to get a better sense of it yourself using some materials that a librarian might be able to provide to you. Do not use chatgpt to try to interpret the law, same thing, and do not try to get chat GBT to apply the law to your situation. These are all risky uses of it, in my personal opinion, and ultimately, for all of us, we are all required to live up to the requirements of the court practice directions and the rules of procedure that govern the same.

55:00

Operating a verifying and certifying of the authorities that we are referring to.

55:06

So I’m going to call it there, but I do have a few more things if we have time. Okay? Liz, thank you. Thank you very much, Annette, and I appreciate your insight on that topic, to verify, verify and verify again, which is key as a legal professional, but even as self reps, verifying your information is key, and like you’ve mentioned, use the can

55:31

lead. Go on there. Look up your case law, look up and then find the correct citation, so you can add that to your documents, which gives you credibility. And I want to thank you, Annette, and we will come back to you. I just wanted to mention a couple of things. There’s been a couple questions in the chat that, what is the LSO rules on AI? And there is information on the LSO website. It’s for general AI use in law. And if you go on the Law Society of Ontario website, and put that in it will direct you to the page that you can go to to find the AI information and the LSO rules on using AI. There’s also been a question about, where are there reputable AI learning? And the one thing is, there is a course coming up that Osgoode is offering. It’s called Critical Thinking for the AI era. It’s happening on May 27 2026 it is between one to 4pm and that is available for anyone that wants to think about, well, how can we use AI responsibly? Because remember, our presentation here today is for general information only, and it is not considered legal advice. And we can’t stress that enough, this is our opinion, and it’s just general information to give those here on the line and those watching this in future, general information on AI. Now I’m going to take it back to Maura and ask her a couple questions about maybe prompting an instruction. So when we think about prompting as a legal skill, is there any kind of tips and tricks you might be able to offer?

57:10

Sure, so these tools, these large language models, are very, very sensitive to prompts.

57:21

So you can change a couple of words in your prompt, and it can completely change the answer you get. You can also ask the same question twice and not get the exact same answer. So I think the best way to think about prompting is you have a really enthusiastic eight year old assistant, and you are going to explain to them what you want. So you aren’t just going to say to an eight year old assistant, go write me a brief on my landlord tenant dispute, you would say to them, I’m having an argument with my landlord. Well, first, you might even give them the context, I’m renting an apartment and I gave my landlord $1,000

58:15

security deposit. My landlord is refusing to give me my $1,000 security deposit back. But you have to. You have to give the system context. It needs to know exactly what it is that you want, who your answers for? Who’s your audience. So if you’re writing for a court, you can say, don’t just say, write me a brief, but say I’m in this particular court. You might even say before this particular judge. And here’s the issue I have, and here’s what I need to draft, and it needs to be 100%

58:58

accurate. So do not cite anything or do not include anything that you are not certain about. And you can in your settings, when you set up chat, GPT or one of these other systems, there’s something called temperature. Higher temperature means be more creative. Lower temperature means be more factual, and you want to make sure your temperature is set low. That’s one thing. You can also give custom instructions so you can say, I don’t want you to try to just please me. I want you always to present both sides of the argument that way. You know you can put that in as a custom instruction. So it’s behind the scenes instruction for everything you always do.

59:59

Or.

1:00:00

Our accuracy is incredibly important to me.

1:00:03

I’d rather you tell me you don’t know where you’re not sure then, then to give me an answer that’s wrong.

1:00:09

You want to break up your tasks into chunks. So the same way you wouldn’t say to the eight year old, you know, go build me a house, or whatever you or go clean the house. You may say, why don’t you start in the kitchen, and I need you to wash the dishes. And here is the dishwashing liquid, and here is a sponge and so on. So you want to give your system detailed prompts and instructions, step by step, what to do. So you might not do the whole brief in one piece. You might say, let’s start with the introduction. Here’s what I want the introduction to say. I wanted to to read professionally for a judge, and the more of these kind of, yeah, I want it to be a page and a half. So the the the more specific and clear and like chunks that you give, the better the output you are going to get, the more vague the

1:01:19

you know, who knows what you get? The same thing when you give the eight year old, a an enthusiastic eight year old, a very

1:01:27

generic description, who knows what they come back with. They’re well intended. They’re trying to please you, but it may not be what you want. So the

1:01:36

the more you specify up front, about tone, about audience, about length, about the format you want it, those are going to be much better prompts. You can also use the system to help you improve the prompts. You can ask Gen AI, how should I phrase this to get a better prompt? Or how would you break this task down into little pieces to make it easier? So and you shouldn’t always assume that it’s just one and done, you will have to iterate with prompts, and you get something back, and you’ll see, no, that’s not exactly what I wanted. Well, then give it feedback and tell it it’s too long. It’s too snarky. I want it to be more serious and professional in tone. I need it to be three paragraphs, not two pages. I want bullet points. I don’t want bullet points again. Break it down into small chunks as much context as you can iterate, and then also ask the tool itself for help. Liz, any other questions? No, that’s the question that I have for you today. And yeah, you know I’m loving this, because I hear building blocks. So when I think of building blocks, I think of when you’re talking about that enthusiastic eight year old, and when I think about LEGO, and in order for you to get that complete structure, you need to start with your building block. Start with the base. And that’s what you’re mentioning when you’re talking about customizing, setting your temperature, custom instructions so that the AI, whichever one you happen to be using, is getting detailed, structured information on how to give you the output that you’re looking for. So thank you for that. I’m learning something here today, which is amazing. So I’m going to move on to a same general topic. Oh, did you want to say something in that? Yeah, I was wondering if you’d be if I could take two seconds to just show people this really interesting way of visualizing the general the generating process. This is a

1:03:50

academic that actually teaches by building little AI applications. So if you can look at this, what he’s giving the students have entered like a legal question into the left hand box on the screen. And if you look on the right, if you if it wasn’t color coded and boxed out, it would look like a summary of a particular legal principle. But what he did is he asked AI to show the tokenization of it so every single box on this screen has been generated. This is the AI saying I’m going to put the number one first, and then a period, and then Smith, and then V and so it’s literally generating every single piece of this answer. It’s not thinking about an answer and putting a nice answer together. It is generating piece by piece by piece, and the color coding is showing you how certain it is about what it’s putting there. So the yellow means it’s not it’s not as certain as the green ones. I’m just showing you this to show you that AI is not thinking. It is actually generating every single piece.

1:05:00

Piece of the answer, which explains how, you know, helps to shore up what more is saying about how it can hallucinate, because it’s not thinking about anything. I just thought that was really interesting little visualization that might be helpful for people.

1:05:13

Thank you. And that reminds me of when you put a question in, like copilot or chat GPT, and you see it doing that circle. I call it the wheel of death, but it isn’t really. It’s just the circle. And that’s what it’s doing. It’s just generating all those tiny little boxes. And like you said, it’s not really thinking in the thought that we are, the thought process that we have. It’s just putting together all these building blocks. And so in order to get the correct structure that we’re looking for, putting those key prompts in is going to help to get the, well, the,

1:05:51

I guess, the output that we’re really looking for, the required output. And even when we get the required output, we still have to verify. We still have to go in and check our case law and use citations to bring up the correct information. So thank you for that demonstration. I’m hoping, because I’m a visual learner as well, that that also helped others to really explain what Maura was talking about, what you’re talking about, essentially what we’re talking about here today on how to use proper AI tools to get the desired results. So I’m going to draw in Lisandra again when we talk about, how can we support self rep litigants so that they can access the correct justice for equity, like, how can we support them from creating more harm.

1:06:48

Lisandra, you’re muted.

1:06:53

I call it the dreaded mute button.

1:06:56

Thank you. So I think one of the things that when you’re talking about self litigants. It’s really focusing on the support that we could give them and and that kind of

1:07:09

talked about some of it earlier, is really giving them the idea or where they can go for research, like she talks about going towards a librarian, a legal librarian, or finding clinics that are legal clinics that offer support. There might be community groups that have legal clinics if you’re not too sure about how to go about creating, creating a case or a brief, or things like that. And again, if you’re really looking at

1:07:42

at using AI like it can help you understand, but it can’t really again, the rhetoric that we keep saying is it doesn’t help you decide what what is professional at the end of the day that is really on you. I would say never submit AI generated court documents. Have we stated before.

1:08:03

It doesn’t know specific facts as a self, you will know the facts. So if you want to organize your facts, like and then talked about the way you should use it, use it to organize your thoughts, organize your facts, but not really to generate and then again, it goes back to the prompting, like Mara. Mara talked about it’s really putting in those criteria is be specific about what you’re looking for. And honestly, if you’re not sure, if you’re unsure, stop and ask. Not the AI, but a real lawyer. I think, if I’m correct, you guys could 30 minutes. It’s usually a free timing to speak with the lawyer.

1:08:44

You get a if you’re not really paying for that, so stop and ask a lawyer get that 30 minute consultation.

1:08:52

Thank you. I’m not sure if Annette might have anything to add to how we can support self rep litigants, so I know that there are other so much out there. But how do we hone in on this? I have a couple more slides that are actually direct help but, and then a couple comments afterwards as well, but I just thought I’d throw this out. So seems like a good time to talk about it. So this came out last Thursday, and so this is now open AI. This is chat GBT has been accused of practicing law without a license in a lawsuit brought by Nippon Life Insurance Company of America. So according to the insurer’s complaint, which was filed on Wednesday in the North District of Illinois, open AI’s artificial intelligence platform chat GBT pushed a woman seeking disability benefits to breach a settlement agreement, fire her lawyer and file dozens of motions that serve no legitimate legal or procedural purpose. So I mean, this is my first starting point, which is, Please don’t fire your lawyer in favor of chat GPT. And second of all, we.

1:10:00

Wouldn’t we would not rely on chatgpt to help us with any aspect of our legal claim other than the kinds of small helps that we’ve talked about here. So this is how bad it can get when people put their faith in these systems. So that’s just a little sort of storytelling on the self represented litigants piece. And I just wanted to show this little slide, which shows you the can Lee Search Plus on the right hand side when so on the right hand side of the screen. When you visit their website again, you can register for free for that my can Lee account to use that search plus

1:10:36

and then, but when you are using it, just remember that it is not summarizing an answer for you. What it is doing is on the right hand side, it shows you the process that it went to. It went through to come up with some authorities for you to review and and my main takeaway here is read that process summary carefully and heed the warnings that are provided. So in this little example, I asked it a question that I knew it could not answer. I asked it, is it is it permissible to have a campfire in a residential area in the city of Windsor? How did I know it could not answer this? Because can Lee does not have city bylaws. City bylaws are not available in Westlaw Lexus or can Lee, and so they will not be able to generate an accurate answer, because these systems can only generate answers off of data that they have access to, they cannot make up for deficient data. So I knew Canley couldn’t answer the question. So if you read this process summary carefully, it specifically says a limitation is that Windsor municipal bylaws themselves may not be fully available on Canley, so results may skew towards cases mentioning the bylaw rather than the bylaw itself. So just a takeaway for you as you are trying to use can lead search. Plus, if you decide to try it, of course, you can always look through the authorities to see what it’s giving you, but be paid careful attention to the the warnings it’s giving you in its process. Summary,

1:12:19

when you are citing authorities, it’s not too difficult today, now that you have can lead just remember that a proper citation always appears across the top of the document. Just use the citation that it gives you. Don’t change it. So this is an example of how to cite the rules of civil procedure. Take it right off the top of the document, okay, and then don’t forget, though, that you also need to tell the judge where to look within the document. So for statutes and regulations, you need to tell them what section and subsections you’re referring to. So in this example, the pinpoint section is section 4.06,

1:13:00

subsection 1d that tells you how to prepare a factum. So this is just my example. And so if you were telling the judge, Judge, please see this section of the rule or the section of the reg or the act, you’re going to add it to your citation, which is shown in yellow at the bottom of this slide. So this is how you would add that to a legislative citation. If you are citing a judicial decision or a case again, take the information off the top of the case, and then you’re going to add the paragraphs that you want the judge to look at, as shown in yellow on the bottom of this slide. So you’re going to take the citation, and then you’re going to add at Perez eight and nine as some examples. So I just thought I would share those as helps for any self represented litigants on the line. And then also, just generally, to mention, again, your law librarians. They cannot give you legal advice, but they can at least show you some books and some articles that you can read to get familiar.

1:14:04

And of course, the National self represented litigants project. And if you’re in other jurisdictions like Alberta, there are organizations called public legal education centers. And so for example, if you go to Alberta, there’s libraries that exist just to help the public, and I think those are going to be my my pointers there. So thank you.

1:14:27

Thank you very much. I just wanted to bring up a point that I know Lisandra mentioned about the 30 minute free consultation. I know there’s been some inquiries about it.

1:14:38

So the reason like, yes, there are still some companies, some legal companies out there that do offer a consultation. However, consultation is not advice. A lawyer or a legal representative cannot give you advice unless we are retained. So what? What it is is a consultation. Now some firms no longer give consultations.

1:15:00

Conversation, because there’s been so much things happening in the industry that are dissing legal representatives from doing that now. So all you have to do is either you go on the website to see if they offer consultation. Some consultations now are paid services because of what’s been happening in the industry, and those that do offer the consultation, you’re more than welcome to contact them, either by phone or by email. Some of them have like a little inquiry box where you put in your information and an email goes directly to the company, and then you can discuss with them. At that time, it’s then up to you as the individual, whether or not you want to move forward with that particular organization, or, you know you have to build a rapport see if you feel comfortable. Do you really trust this person? As we know that not everything is for everyone, so sometimes you’re not really sure of that relationship. Maybe using a referral might be best as well. I’m not 100% sure. Now, for those of us, can I just add one thing to what you’re saying,

1:16:05

if you can afford a half hour or an hour of an attorney’s time, do your work using

1:16:15

the chat bot and then bring it to the Attorney and spend the one hour, very well spent to have them check your work. That’s a really good sort of hybrid. Instead of having them do all the work and spending 20 hours, you do the work and then bring it to a professional or bring it to an organization that offers, maybe they offer a limited amount of time, for either free or for a limited amount of money, and then have the professional check your work. That’s all I had to add. Amazing. Thank you so much for that point. This is why we have a panel, so that you know we’ve got, what, four people here using their heads instead of just one, and when one point comes out, someone else backs it up. So I want to thank the panel so much for your input. Now there’s also a question of, well, where do self reps go? Because remember, self reps, they may not have a lot of funds. How do they loan? About how do they learn? Sorry, about appropriate AI, if some of you are students, I know that there’s so many colleges and universities now offering these AI courses because of what is happening in the industry with AI. I know that Toronto metropolitan university is creating some AI literacy courses. I know that u of t is doing the same thing, and it’s, it’s almost like a domino effect, because AI is becoming so popular, and it what, you know, everyone has that mobile device in their hand, which is like a handheld computer that we all walk around with. And so if you happen to be a student, and even if you’re not a student, you could probably take a part time night course that is going to allow you to learn about proper AI literacy, and it’s it’s going to be beneficial, it’s going to be helpful. And then coming to webinars like this, where we’re giving you general information on how to move forward, to know, how do I use AI safely? All right, now we’re going to go into some rapid questions here, and I’ve got a really big one for the entire panel, and I’m going to ask you guys to go one at a time. If you could give one piece of advice to someone using AI for legal research tomorrow. What would that be?

1:18:35

I see Mara, you’re unmuted. So did you want to start with that? I would agree with my colleagues, and I would say, first stop Canley, not first stop chat, even though you think chats easier, first stop Canley,

1:18:55

yep. 100% I was going to say, don’t use it for legal research. But can lead Yes,

1:19:02

and I’m going to, again, apply what my two colleagues have said, and from all intents and purposes, the research that I’ve been doing, it took me to Canley, and not really a legal background, but I’ve learned so much from just applying my educational aspect to the law perspective, and I would say definitely can lead is your best bet.

1:19:26

Thank you. That works for Canada. Obviously, if you’re in the United States, you got a problem because the law isn’t available for free in the US.

1:19:35

Mara, I love you.

1:19:37

Now, when we think about practical steps for users, right? What do we think of now, I know we’ve mentioned so much here today, but when we think about self reps, even licensed individuals, some practical steps of how to properly use AI.

1:19:55

I know Annette mentioned a couple. I know Mars mentioned some. Lisandra. Now this is just.

1:20:00

Just rapid fire questions. So maybe you could give one each.

1:20:05

I’ll start spend the $20

1:20:07

if you can afford it, and see what it can do and what it can’t do. Start with, don’t put anything confidential in, as Lisandra said, you don’t want to be putting your health information and that kind of stuff in. But see what it can do help it avid plant look, take a picture of what’s in your refrigerator, and ask it to help you to plan a menu for dinner. Ask it to help you to plan your next vacation. And get a sense of what the tools do well and what they don’t. They have what’s called a jagged edge. They do some things remarkably well, and some things they’re just terrible at, and you got to get a sense of that so that when you’re using them in legal work, if you’re going to use them that way, that you have a good sense of what they can do and what they can’t do.

1:20:53

I would say, in addition to that, I would say, Remember again, technology enhances your work. It does not give you. It does not become you. You still have to be the human in the loop, as we call it, in the technology world. And when it comes to AI, a lot of biases do exist.

1:21:15

Try it out first, as Mara said, like even if you don’t have the $20 and you want to, I know, student budget and both. So if you don’t have the $20 and you just want to use the basic I’ve used it to create recipes with things that I have in my fridge that I didn’t know, go together and it came out amazing, like I’m a chef. So I would say definitely use that. Another thing is trying to start learning what Mara said about prompt prompt generation. We’re not expecting it to become coders, but that prompting is very important be specific with what you’re asking it.

1:21:50

And I would say on that point, on the to on the prompting part as well. Just remember that AI cannot handle a huge amount of text that you’re putting in and generating a huge amount of text that you’re taking, that you’re receiving from it, because it all has a it has a limit to how much it can take and how much it can give out in a in a specific interaction. It’s called a context window. So what I tell people is break everything down into very small pieces. So if you have a little question for it, put the question in, get in, get its output, and then start a new conversation with it, and just keep everything very clean,

1:22:31

amazing. Thank you to our panel. You guys been amazing here today. I really appreciate your insight on AI. You guys have taught me so much here today, because I’m what you call an AI, and I hope I don’t offend anyone by saying an AI virgin, because I like, I mean, I don’t even know what I’m doing, and people laugh at me all the time, and I’m like, Okay, I’m, you know, and I’m not going to date myself, but telling you how old I am, and you know what my era was, but using AI for me as a foreign thing. So I want to thank you guys so much for the education that you’ve given me here today.

1:23:06

And I just want to also reiterate that there’s been questions about whether or not this has been accredited for CPD hours, and we’re working on that, and so it will be on the website once we get the information from the LSO, and for those that are here today, it’ll go out by email so that you guys will know what the accreditation is. Now this is an hour and a half webinar, so that is what we’re hoping to get from the LSO, but we haven’t heard back yet. So for those of us wondering, we will be sending out that information as soon as it becomes available, I’m like going to turn everything over now to Dana to wrap it up so that we are on time. Thank you to the panelists. Very, very much. Thanks, Liz, and I have to apologize to you. I was looking at too many documents when I was doing my intro, and I I gave you the wrong last name. Your last name is Smith, of course, not con live, that’s lissandra’s last name. So I apologize. It’s all good. It’s all good. I appreciate that. Thank you so much. We have the same hair color. It’s okay. Thank you. I appreciate both of your grades. Thank you.

1:24:15

So I want to thank Liz certainly. I think that was just a masterful job of moderating what is a really complicated topic and issue, and Lisandra, Moira and Annette, we can’t thank you enough for all of your expertise. You’ve been so generous with your time, both leading up to today and today. So thank you all so much. We’re also really excited about this partnership that we have between the NSLP and the oabp, and we’re certainly hopeful that we might be able to continue that in the future. So we’re grateful to both organizations for everybody’s work in preparing for today.

1:24:56

A little bit more housekeeping. The recording will be available.

1:25:00

Bowl, certainly through the NSRLP, I think I can probably confidently say that the OIB will make it available as well through their either website or social media or, you know, whatever platforms. Liz is not in great so the recording of this as soon as we can, we can get it processed and and up. We will make sure we do that, and that’ll be shared across our social media platforms and newsletters and and all of that. So we’re finishing with just a few minutes to spare, which I’m sure everybody appreciates. On a Friday, I want to thank all of our attendees for spending the time to

1:25:35

to be with us today, and certainly I, like Liz was saying, I’ve learned a lot today. It’s been a very informative session, and we hope, certainly NSRLP, I can speak from our perspective, we hope to put out more content in the future on the effective and safe and appropriate use of AI and Liz, I would guess that you’re feeling the same 100%

1:25:58

and on behalf of the Ontario Association of Black paralegals. I want to thank everyone for being here today. I want to thank the wonderful panelists for all of their insight and information. And Dana, I want to thank you and your organization as well for partnering with us to bring out this amazing information that is key for what is happening in 2026 and moving forward. So thank you very much on behalf of the organization. Thank you all right, thanks everybody, and have a wonderful rest of your day and a wonderful weekend. Take.