1 00:00:12,870 --> 00:00:14,860 Welcome to studying Law Around the World. 2 00:00:15,010 --> 00:00:17,860 I'm Claudio Claus. In each episode, I talk with 3 00:00:17,870 --> 00:00:20,840 lawyers, law students, and professors from different parts 4 00:00:20,850 --> 00:00:24,740 of the world to talk about legal education, careers, and what the 5 00:00:24,750 --> 00:00:26,420 profession looks like in real life. 6 00:00:26,430 --> 00:00:28,680 We talk about the hard parts, the surprises. 7 00:00:28,690 --> 00:00:30,780 And the decisions that. Shaped their paths. 8 00:00:31,030 --> 00:00:33,920 Whether you're planning to study abroad, thinking about working 9 00:00:33,930 --> 00:00:37,380 in another legal system, or just curious about how law works 10 00:00:37,450 --> 00:00:40,180 around the world, this podcast is for you. 11 00:00:40,190 --> 00:00:43,360 Today I'm interviewing on Macintosh Zeng. 12 00:00:43,470 --> 00:00:47,590 He is returning to the podcast for some new exciting updates. 13 00:00:47,660 --> 00:00:51,430 We first talked about court ready, I believe last year and 14 00:00:51,440 --> 00:00:54,830 now we have some very, very interesting topics to to be 15 00:00:54,840 --> 00:00:57,240 discussing in this episode. Thank you so much for making the 16 00:00:57,250 --> 00:00:59,330 time to be here, Don. Well, thank you very much for 17 00:00:59,340 --> 00:01:01,190 having me, as always, glad to be back. 18 00:01:01,240 --> 00:01:03,830 Amazing. And she she started podcast. 19 00:01:03,840 --> 00:01:06,930 I always appreciate if you can just introduce yourself briefly 20 00:01:06,940 --> 00:01:09,550 for maybe listeners who are listening to these episode 21 00:01:09,560 --> 00:01:13,390 before our previous ones or who are kind of first coming to the 22 00:01:13,400 --> 00:01:15,110 podcast. Yeah, absolutely. 23 00:01:15,120 --> 00:01:20,090 So my name is Tom McIntosh Zane. I'm the co-founder of Court 24 00:01:20,100 --> 00:01:23,810 ready.ca, where we built practical tools to help 25 00:01:23,820 --> 00:01:27,950 Canadians navigate the legal system and from time to time we 26 00:01:27,960 --> 00:01:32,020 conduct studies into access to justice issues in Canada. 27 00:01:32,110 --> 00:01:36,300 I used to be a commercial barrister in practicing in 28 00:01:36,310 --> 00:01:39,240 Toronto and I'm also qualified as a solicitor in English, 29 00:01:39,310 --> 00:01:41,840 England and Wales. Amazing, fantastic. 30 00:01:41,910 --> 00:01:45,260 And you kind of get right on the topic of our conversation today, 31 00:01:45,310 --> 00:01:48,800 talking a little bit about court ready and a little bit about, 32 00:01:48,810 --> 00:01:51,560 you know, the whole world of technology, AI and so forth. 33 00:01:51,630 --> 00:01:53,960 It was interesting that in our first conversation we talked 34 00:01:53,970 --> 00:01:56,780 about how there were so many different tools that could be 35 00:01:56,790 --> 00:02:00,290 used in legal practice that actually didn't even need AI. 36 00:02:00,300 --> 00:02:02,940 They were simpler tools, but they just could really 37 00:02:02,950 --> 00:02:06,100 complement the practice. Ever since then, lots, lots has 38 00:02:06,110 --> 00:02:10,520 have have happened and everyone you know from junior lawyers, 39 00:02:10,740 --> 00:02:13,710 self represented litigants and, and even members of the 40 00:02:13,720 --> 00:02:18,810 judiciary are learning about generative AI and maybe using it 41 00:02:18,820 --> 00:02:22,710 to some extent to save time, to save fees and, and to get an 42 00:02:22,720 --> 00:02:24,390 understanding of different things. 43 00:02:24,400 --> 00:02:26,530 But there are some problems with that, right? 44 00:02:26,540 --> 00:02:29,930 So we did learn that there is made-up citations, some kinds of 45 00:02:29,940 --> 00:02:32,850 hallucinations. And specifically the research 46 00:02:32,860 --> 00:02:37,210 you conducted found a great deal, I think over 200 made-up 47 00:02:37,220 --> 00:02:41,450 citations that affected over 100 cases or 100 decisions. 48 00:02:41,520 --> 00:02:44,910 So I, I wanted to hear a little bit about that about, you know, 49 00:02:44,920 --> 00:02:49,780 your, your research how that came about and maybe even are 50 00:02:49,790 --> 00:02:53,190 those the result of AI that was used in the past and now AI is 51 00:02:53,200 --> 00:02:56,510 better and, you know, doesn't hallucinate as much anymore. 52 00:02:56,520 --> 00:03:00,270 What's your view on all of that? Right, So to start with, many of 53 00:03:00,280 --> 00:03:05,250 us know that when we use the publicly available AI tools, 54 00:03:05,300 --> 00:03:09,010 sometimes the tool has a tendency to hallucinate. 55 00:03:09,080 --> 00:03:13,750 So what I mean by that is you ask it a question and it would 56 00:03:13,760 --> 00:03:18,950 provide the answer and the answer sometimes is incorrect. 57 00:03:19,000 --> 00:03:23,090 That happens also in the field of law, right? 58 00:03:23,100 --> 00:03:27,730 So as a as an experiment myself, sometimes I would ask it to give 59 00:03:27,740 --> 00:03:30,630 me the answer to a legal question. 60 00:03:30,640 --> 00:03:34,360 Now for simple ones, it has the tendency to get the answer 61 00:03:34,430 --> 00:03:37,220 correct. But say if you asked any of 62 00:03:37,230 --> 00:03:42,640 these tools that do not have access to a law database, what 63 00:03:42,650 --> 00:03:48,090 are the top five cases in Canada on a particular topic when you 64 00:03:48,100 --> 00:03:51,280 ask that question and you get the answer, I've seen that it 65 00:03:51,290 --> 00:03:54,180 might give me cases that don't exist. 66 00:03:54,250 --> 00:04:00,400 It might give me cases that has a real name, but the substance 67 00:04:00,410 --> 00:04:02,870 of the case is completely different. 68 00:04:02,930 --> 00:04:06,200 Or it would give me quotes from from a case, but then those 69 00:04:06,210 --> 00:04:09,600 quotes actually do not exist in that case itself. 70 00:04:09,660 --> 00:04:14,640 So then I asked myself the question is, has this particular 71 00:04:14,650 --> 00:04:19,560 issue of AI hallucination impacting our courts and 72 00:04:19,570 --> 00:04:23,170 tribunals because we know that a lot of folks who may not have 73 00:04:23,180 --> 00:04:27,480 the means to afford a lawyer or who may not be comfortable 74 00:04:27,550 --> 00:04:33,060 writing in legal English and French turn to these AI tools to 75 00:04:33,070 --> 00:04:36,440 help them write and help them research us. 76 00:04:36,490 --> 00:04:40,160 Other than say a small disclaimer at the bottom of 77 00:04:40,170 --> 00:04:43,360 every tool say you know, blah blah blah can make mistakes. 78 00:04:43,410 --> 00:04:46,040 The tools don't really prompt you that pay. 79 00:04:46,050 --> 00:04:48,900 Wait a second, These cases may not be real. 80 00:04:48,970 --> 00:04:53,060 So for someone without a legal background, when they looked at, 81 00:04:53,120 --> 00:04:58,270 when they see a confident case citation from one of these AI 82 00:04:58,280 --> 00:05:00,060 tools, they think that it is real. 83 00:05:00,130 --> 00:05:03,120 So as the context for why we did this research. 84 00:05:03,190 --> 00:05:08,710 And what was very surprising to me is that since 2024, there 85 00:05:08,720 --> 00:05:13,820 have been 36 court cases on tribunals where the court or the 86 00:05:13,830 --> 00:05:19,300 tribunal flagged that a party submissions contain fictitious 87 00:05:19,350 --> 00:05:23,680 citations, 136 decisions. So that's where the court caught 88 00:05:23,690 --> 00:05:26,980 people submitting cases that don't exist. 89 00:05:27,050 --> 00:05:33,400 There are separately 18 other cases where the court warned a 90 00:05:33,410 --> 00:05:39,810 party about AI misuse. But looking at the 136, they are 91 00:05:39,820 --> 00:05:43,750 all over the country. They're in 45 courts and 92 00:05:43,760 --> 00:05:46,850 tribunals. And in total, there are at least 93 00:05:46,860 --> 00:05:52,290 247 fictitious, fictitious citations that were submitted 94 00:05:52,380 --> 00:05:56,850 across the 136 decisions. And the trend is accelerating. 95 00:05:56,920 --> 00:06:03,140 We had about 7 or 8 instances IN20002480 cases in 2025. 96 00:06:03,200 --> 00:06:09,070 And in the first quarter of 2026 alone, there were 50 cases. 97 00:06:09,360 --> 00:06:14,990 So naturally the question is these were the instances that 98 00:06:15,000 --> 00:06:17,540 were caught by the courts and tribunal. 99 00:06:18,280 --> 00:06:21,770 But that means statistically, and again, I am speculating here 100 00:06:21,780 --> 00:06:25,330 a little bit, there's there ought to be instances where 101 00:06:25,340 --> 00:06:30,900 these cases made into an actual decision unnoticed by everyone 102 00:06:30,910 --> 00:06:33,710 along the way. Now, we actually have seen the 103 00:06:33,720 --> 00:06:37,450 first instance of this being reported where a decision 104 00:06:37,460 --> 00:06:40,390 contained fictitious citations is a case out in Quebec. 105 00:06:40,400 --> 00:06:44,170 Obviously we don't know who was ultimately responsible, but that 106 00:06:44,180 --> 00:06:49,170 was the first at least reported case of a decision containing AI 107 00:06:49,180 --> 00:06:53,690 hallucinated citations that no one was able to stop before it 108 00:06:53,700 --> 00:06:55,850 was released. Well, thank you so much for for 109 00:06:55,860 --> 00:06:57,740 explaining and and going over that. 110 00:06:57,980 --> 00:07:01,620 I find it interesting that that you see this jump and and I'm 111 00:07:01,630 --> 00:07:03,520 always wondering different, different things. 112 00:07:03,530 --> 00:07:07,100 So I wonder if actually when the citation and and the decision 113 00:07:07,110 --> 00:07:10,500 maybe came about, probably those were written quite a quite a 114 00:07:10,510 --> 00:07:12,200 while before that happened, right? 115 00:07:12,210 --> 00:07:16,140 So I mean, imagining the faction was written probably even a year 116 00:07:16,150 --> 00:07:19,600 before actually a decision came about. 117 00:07:19,690 --> 00:07:22,690 I'm curious to understand as well as as as far as you know, 118 00:07:22,700 --> 00:07:25,680 what the tools goes. And I really appreciate you 119 00:07:25,690 --> 00:07:30,280 mentioning this is specifically that, you know, the the database 120 00:07:30,330 --> 00:07:33,550 mentions, you know, the the court citations where judges 121 00:07:33,560 --> 00:07:37,340 actually noticed and flagged in, but then there could be things 122 00:07:37,350 --> 00:07:40,280 slipping under the radar, the radar. 123 00:07:40,290 --> 00:07:44,870 So what's really happening behind the scenes with all of 124 00:07:44,880 --> 00:07:48,930 these fake cases? So you're right in that there is 125 00:07:48,940 --> 00:07:51,670 a delay factor, right? Because sometimes you would 126 00:07:51,680 --> 00:07:57,080 write a factum or any submissions and that's gets 127 00:07:57,090 --> 00:08:00,610 submitted and of course decision isn't released a while later. 128 00:08:00,680 --> 00:08:04,890 So even though these decisions are coming on 2025 and 2026, we 129 00:08:04,900 --> 00:08:08,550 are probably looking at the submissions that were happening 130 00:08:08,560 --> 00:08:13,630 six months to a year ago. And so then the next question 131 00:08:13,640 --> 00:08:16,070 is, is it going to get worse or is it going to get better? 132 00:08:16,080 --> 00:08:18,470 And, and that's what we're trying to figure out whether or 133 00:08:18,480 --> 00:08:22,800 not because the tools themselves are getting better that we work. 134 00:08:22,810 --> 00:08:27,390 We, we hope to see that this trend decreasing, although from 135 00:08:27,400 --> 00:08:30,330 what we've observed, there's just more and more decisions 136 00:08:30,340 --> 00:08:32,690 coming out. And also because of the way that 137 00:08:32,700 --> 00:08:35,730 we're conducting our research, which is primarily relies on 138 00:08:35,740 --> 00:08:39,270 reported decision. As a lawyer myself, I know often 139 00:08:39,559 --> 00:08:43,429 the decisions are not reported or published or made their way 140 00:08:43,440 --> 00:08:45,830 to county. So we might just be looking at 141 00:08:45,840 --> 00:08:48,110 the tip of the iceberg. Here, very, very interesting. 142 00:08:48,120 --> 00:08:51,010 Well, thank you for for sharing those details. 143 00:08:51,080 --> 00:08:54,770 So specifically speaking on legal citations, and I think 144 00:08:54,780 --> 00:08:57,450 this might be important even for people who are outside of 145 00:08:57,460 --> 00:08:59,650 Canada. I understand that our Canadian 146 00:08:59,660 --> 00:09:03,430 legal citations, well, maybe in the US to some extent, but they 147 00:09:03,440 --> 00:09:06,150 follow very strict predictable format. 148 00:09:06,240 --> 00:09:09,570 So you've shown that that that makes sense and makes it easy 149 00:09:09,580 --> 00:09:13,010 for AI to kind of mimic that structure, right. 150 00:09:13,060 --> 00:09:16,750 So why does that structure sometimes will trick people or 151 00:09:16,760 --> 00:09:19,390 catch, you know, lawyers off guard? 152 00:09:19,450 --> 00:09:20,860 Yeah. So that's a great question. 153 00:09:20,870 --> 00:09:24,200 I think for the folks who may not be as familiar with the 154 00:09:24,210 --> 00:09:27,260 Canadian system as you are. We obviously like the other 155 00:09:27,270 --> 00:09:31,120 common law jurisdictions, rely on cases as president. 156 00:09:31,130 --> 00:09:35,260 So if a judge has decided a case in a certain way and you have a 157 00:09:35,270 --> 00:09:39,360 similar case coming up next, you would want to cite the results 158 00:09:39,370 --> 00:09:43,560 or the law as it was interpreted in the prior case to support 159 00:09:43,630 --> 00:09:47,340 your argument. Now, Canadian case citations all 160 00:09:47,350 --> 00:09:50,860 tend to follow what's called a neutral citation pattern, which 161 00:09:50,870 --> 00:09:57,020 is party V versus party B, comma, the year, the court 162 00:09:57,030 --> 00:10:02,600 abbreviation and then a naturally increasing integer. 163 00:10:02,670 --> 00:10:08,730 So a case could be you know from V Jerry 2021 Glen SC1234. 164 00:10:08,740 --> 00:10:13,390 So you know just by that it is a two parties Tom and Jerry and 165 00:10:13,400 --> 00:10:17,490 the case and whatever the year and OSC stands for Ontario 166 00:10:17,500 --> 00:10:22,350 Superior Court. This system makes it easy for an 167 00:10:22,420 --> 00:10:26,230 AI to mimic. And So what I mean by that is 168 00:10:27,080 --> 00:10:30,910 we, we tend to use sort of moral terms to describe, you know, 169 00:10:30,920 --> 00:10:34,470 lying or making things up. AI, as many of us know it 170 00:10:34,480 --> 00:10:39,870 predicts what's the most likely answer based on the prompts that 171 00:10:39,880 --> 00:10:41,970 you give. So let's use an example. 172 00:10:42,040 --> 00:10:47,490 If you ask an AI what is the, the, the most important case in 173 00:10:47,500 --> 00:10:52,150 Ontario that talks about spousal support in the context of family 174 00:10:52,160 --> 00:10:55,870 law, you ask that question, if the AI does not have access to a 175 00:10:55,880 --> 00:11:00,030 database, it's going to try to figure out what is the most 176 00:11:00,040 --> 00:11:03,050 likely answer to that based on patterns. 177 00:11:03,060 --> 00:11:06,050 So it's a family case, right? And so in Canada, the most 178 00:11:06,060 --> 00:11:10,550 common last name is Smith. So Smith V Smith is the most 179 00:11:10,690 --> 00:11:13,320 likely answer for an Ontario case. 180 00:11:13,330 --> 00:11:17,300 The chances are it's going to be out of the Superior Court. 181 00:11:17,310 --> 00:11:21,540 So OSC is the the, the court abbreviation. 182 00:11:21,610 --> 00:11:28,400 So Smith V Smith 2017 OSC 135. Then that's how the AI would 183 00:11:28,410 --> 00:11:32,600 produce a case that really looks real but isn't because it is 184 00:11:32,610 --> 00:11:35,860 predicting the most likely outcome based on the prompts 185 00:11:35,870 --> 00:11:38,130 that you give. And you know, the second part of 186 00:11:38,140 --> 00:11:41,240 your question is why does this structure trick people and why 187 00:11:41,250 --> 00:11:43,000 does it cache so many junior lawyers? 188 00:11:43,010 --> 00:11:48,150 Off guard is when I used to practice, we were always, it was 189 00:11:48,160 --> 00:11:51,560 important to cite as many cases as possible to support your 190 00:11:51,570 --> 00:11:55,640 argument, have a legal position. And then in your footnote, you 191 00:11:55,650 --> 00:11:58,090 want, you might have 5 or 6 different cases. 192 00:11:58,100 --> 00:12:01,810 So, you know, the Ontario courts have always ruled in this way. 193 00:12:01,880 --> 00:12:05,690 See comma, for example, case 123-4567. 194 00:12:05,800 --> 00:12:10,670 There was a time where you could trust that those cases are real. 195 00:12:10,680 --> 00:12:14,670 Now they may not actually stand for the arguments the other side 196 00:12:14,680 --> 00:12:17,850 claims that if they stand for, but you can trust that these 197 00:12:17,860 --> 00:12:20,390 cases actually exist. That's no longer the case 198 00:12:20,400 --> 00:12:24,590 anymore because even though it will look real and it can even 199 00:12:24,600 --> 00:12:27,930 trick lawyers because of the fact that it follows A neutral 200 00:12:27,940 --> 00:12:31,810 citation pattern, they could be actual cases. 201 00:12:31,820 --> 00:12:34,650 And one way I can demonstrate it is that I'm going to, you know, 202 00:12:34,700 --> 00:12:37,880 show my screen just for a little bit, but to see what some of 203 00:12:37,890 --> 00:12:40,420 these fictitious citations look like. 204 00:12:40,430 --> 00:12:42,450 So you see here, I'm going to zoom in just a little bit. 205 00:12:42,520 --> 00:12:46,570 This is on court ready.ca where you can see this fictitious 206 00:12:46,580 --> 00:12:50,260 citations in Canadian courts. So what we have on the left hand 207 00:12:50,270 --> 00:12:52,080 side, actually let's do this quiz. 208 00:12:52,130 --> 00:12:55,060 Which one of these is a fake citation? 209 00:12:55,130 --> 00:12:57,570 Which one, Claudio, you pick? Well, I would say the one that 210 00:12:57,580 --> 00:13:02,820 repeats the same last name twice just just for just for that 211 00:13:02,830 --> 00:13:04,370 case. But maybe the the lack of the 212 00:13:04,380 --> 00:13:07,110 period after V makes me doubt my on. 213 00:13:07,160 --> 00:13:09,210 So let's let's go with Viper. Viper. 214 00:13:09,280 --> 00:13:13,170 OK, Piper V Piper was a real case and Brown V Swanson is a 215 00:13:13,180 --> 00:13:15,690 fake case, right? And again, to me, like if I 216 00:13:15,700 --> 00:13:19,520 didn't know British Columbia, BC S is the Supreme Court of 217 00:13:19,530 --> 00:13:22,170 British Columbia out in the western part of Canada and let 218 00:13:22,180 --> 00:13:25,230 unless I check, I would have not known that one of them is fake. 219 00:13:25,280 --> 00:13:28,590 So let's do one more round. Which one of these is a fake, 220 00:13:28,600 --> 00:13:31,030 fake citation? Claudio, your your turn. 221 00:13:31,220 --> 00:13:36,670 Let's go with Wallace V Crates Marine Sales. 222 00:13:37,120 --> 00:13:39,330 Already they're both fake, right? 223 00:13:39,340 --> 00:13:43,210 So this again, like they both look very real and I'm pretty 224 00:13:43,220 --> 00:13:47,680 sure there's like a similar case that's that's about Valgardson. 225 00:13:47,690 --> 00:13:50,450 Alright, so the final round, let's see you pick. 226 00:13:50,540 --> 00:13:53,730 Yeah, I mean, it's just a fun guessing game here, but let's go 227 00:13:53,740 --> 00:13:57,730 with Alessandra V Briggs. OK, so they're both real, right? 228 00:13:57,740 --> 00:14:00,460 And those were actually two of my cases that I did when I used 229 00:14:00,470 --> 00:14:03,950 to practice. So this goes to show that you 230 00:14:03,960 --> 00:14:07,490 can't really tell if if lawyers can't tell, then self 231 00:14:07,500 --> 00:14:11,790 represented litigants who may not even know about how cases 232 00:14:11,800 --> 00:14:15,430 work or how case citations working in Canada, they want to 233 00:14:15,440 --> 00:14:18,890 be able to tell judges and clerks again, they're so busy 234 00:14:18,900 --> 00:14:24,530 already and they rely on lawyers or self reps providing cases 235 00:14:24,640 --> 00:14:26,700 that are real. And we almost take it for 236 00:14:26,710 --> 00:14:29,360 granted, like, frankly speaking, you know, some of this research, 237 00:14:29,410 --> 00:14:31,900 this wasn't an issue before 2020. 238 00:14:31,910 --> 00:14:36,020 Four, no, very interesting. I, I also see and I think all of 239 00:14:36,030 --> 00:14:39,040 the legal community to some extent have been following the 240 00:14:39,050 --> 00:14:41,340 difference, you know, decisions about that. 241 00:14:41,350 --> 00:14:44,220 A lot of them, you know, end up making it to the to the 242 00:14:44,230 --> 00:14:48,690 headlines. It seems like in the beginning 243 00:14:48,700 --> 00:14:51,350 and in in a few different jurisdictions, parts were kind 244 00:14:51,360 --> 00:14:55,530 of giving a warning at 1st and then now we have documented 245 00:14:55,540 --> 00:14:59,610 consequences or that, you know, costs, whether that be the case, 246 00:14:59,620 --> 00:15:03,830 gets referred to the regulator. So for lawyers watching this 247 00:15:03,840 --> 00:15:07,870 unfold, what are some of these maybe let's call it an unwritten 248 00:15:07,880 --> 00:15:11,690 rules since I guess we haven't codified anything like that, 249 00:15:11,740 --> 00:15:15,270 that judges are following when they see a I misuse. 250 00:15:15,320 --> 00:15:18,470 I think there's a there's a possibility of using some kind 251 00:15:18,480 --> 00:15:23,130 of precedent saying that you're maybe misguiding the courts or, 252 00:15:23,140 --> 00:15:26,310 or your other person. But I'm curious, you know, how 253 00:15:26,320 --> 00:15:30,150 the this risk landscape changes in the profession when we see 254 00:15:30,160 --> 00:15:32,850 these this new phenomena happening? 255 00:15:32,920 --> 00:15:36,910 Well, the risk landscape changes dramatically because the courts 256 00:15:36,920 --> 00:15:39,210 and tribunals are getting pretty fed up. 257 00:15:39,360 --> 00:15:43,630 And you are very much accurate in that in the beginning stages, 258 00:15:44,020 --> 00:15:48,460 the courts were almost, well, they're not almost they are 259 00:15:48,470 --> 00:15:52,130 giving benefit of the doubt that those who are providing these 260 00:15:52,140 --> 00:15:55,790 citations do so in good faith and they're not misleading the 261 00:15:55,800 --> 00:15:58,570 court. And secondly, what I think was 262 00:15:58,580 --> 00:16:01,980 happening at the time is that many people didn't even know 263 00:16:01,990 --> 00:16:05,690 that these are AI generated citations. 264 00:16:05,700 --> 00:16:08,670 So I'm just kind of positive for one second to, to, to discuss 265 00:16:08,680 --> 00:16:12,940 this is that in our research, we found cases which we didn't put 266 00:16:12,950 --> 00:16:17,780 in the database that in my own experience, I could tell that 267 00:16:17,830 --> 00:16:25,980 the cases were either hallucinated by AI or had AI's 268 00:16:25,990 --> 00:16:30,100 hand in, in drafting. But those while those citations 269 00:16:30,110 --> 00:16:34,110 were picked up by the judge or the tribunal member, the judge 270 00:16:34,120 --> 00:16:38,750 and tribunal didn't really go the full way by saying you use 271 00:16:38,760 --> 00:16:45,920 AI or, or got into the weeds on how and why these cases came 272 00:16:45,930 --> 00:16:48,120 into your submissions to begin with. 273 00:16:48,130 --> 00:16:51,880 So we are still seeing some instances where the judge 274 00:16:51,890 --> 00:16:56,760 doesn't really know or didn't really identify this being an AI 275 00:16:56,770 --> 00:16:58,820 related issue. But to your point about 276 00:16:58,830 --> 00:17:03,080 sanctions, well we have separately published a database 277 00:17:03,090 --> 00:17:07,849 of sanctions and and this is the show how the courts have reacted 278 00:17:07,859 --> 00:17:10,369 to this this issue. And so I'm going to share my 279 00:17:10,380 --> 00:17:13,990 screen again. This is also part of our court 280 00:17:14,000 --> 00:17:17,890 ready.ca database is on our homepage is like how Canadian 281 00:17:17,900 --> 00:17:22,180 courts and tribunals are responding to AI hallucinated 282 00:17:22,190 --> 00:17:23,430 case law. I miss you. 283 00:17:23,440 --> 00:17:26,310 So I'll give you some of the top line data, right. 284 00:17:26,319 --> 00:17:37,160 So in in this case we have about 42 + 170 a 147 total decisions 285 00:17:37,170 --> 00:17:39,940 that includes some of the AI misuse from our separate 286 00:17:39,950 --> 00:17:41,880 database. And sometimes you know the 287 00:17:41,890 --> 00:17:44,880 different parties appear multiple times, we group them 288 00:17:44,890 --> 00:17:47,950 into one just to not to inflate the numbers artificially. 289 00:17:48,020 --> 00:17:52,330 So the 147 decisions where the court had an opportunity to 290 00:17:52,340 --> 00:17:57,570 sanction 31%, no record of sanctions record just said, OK, 291 00:17:57,580 --> 00:18:02,390 I recognize that you've had fictitious citations in your 292 00:18:02,400 --> 00:18:04,650 submissions. I'm not going to consider them, 293 00:18:05,060 --> 00:18:08,260 right? 49% warnings. 294 00:18:08,270 --> 00:18:11,540 So the warnings will look like, hey, look, you're doing this, 295 00:18:11,590 --> 00:18:15,440 you have to look at our practice direction and you cannot do it 296 00:18:15,450 --> 00:18:18,640 again kind of thing or threatening of further sanctions 297 00:18:18,650 --> 00:18:20,680 of the issue repeats itself. 20%. 298 00:18:20,690 --> 00:18:24,380 So 30 cases, serious sanctions, when I say serious, they include 299 00:18:24,390 --> 00:18:28,880 imposing costs, regulatory referrals, contempt of court or 300 00:18:28,890 --> 00:18:31,660 this is not an interesting one where if you're the successful 301 00:18:31,670 --> 00:18:35,120 party, you are normally entitled to cost. 302 00:18:35,130 --> 00:18:38,240 So other side paying your legal fees, But the court said no 303 00:18:38,290 --> 00:18:42,240 because you used AI and you mislead the court in some way. 304 00:18:42,250 --> 00:18:45,280 So we see that as well. And when you look at trends, 305 00:18:45,450 --> 00:18:46,900 right? So this is the second graph. 306 00:18:46,910 --> 00:18:48,950 OK. So so you look at first of all 307 00:18:49,090 --> 00:18:52,830 an increase in the number of decisions dealing with this 308 00:18:52,840 --> 00:18:55,150 issue, right. So we had you know from one in 309 00:18:55,160 --> 00:18:59,390 the first quarter of 2024 and to let's see this is yeah, almost 310 00:18:59,400 --> 00:19:03,690 50 in the first quarter of 2026. And you see how you're seeing 311 00:19:03,700 --> 00:19:08,550 more serious sanctions and you're seeing more warnings. 312 00:19:08,560 --> 00:19:14,010 And was also interesting about the breakdown is that as a as 313 00:19:14,020 --> 00:19:17,730 expected, right, As a proportion of the total number of 314 00:19:17,740 --> 00:19:20,890 sanctions, those with represented parties, legals, 315 00:19:20,900 --> 00:19:24,830 lawyers and paralegals tend to have a higher percentage of 316 00:19:24,840 --> 00:19:30,070 serious sanctions. 46% of represented parties have serious 317 00:19:30,080 --> 00:19:32,650 sanctions where that proportion is only 15%. 318 00:19:32,660 --> 00:19:34,830 And that makes sense. Self represented individuals, 319 00:19:34,930 --> 00:19:38,740 they're not represented course tend to expect more from 320 00:19:38,750 --> 00:19:41,800 represented parties. So if you look at what kinds of 321 00:19:41,810 --> 00:19:44,070 decision, right. So here we have a, a tabulation 322 00:19:44,080 --> 00:19:46,880 of costs. So thus far the courts have 323 00:19:46,930 --> 00:19:52,200 ordered at least $30,000 Canadian of sanctions for 324 00:19:52,250 --> 00:19:55,340 improper AI use. And yeah, so if you look at the 325 00:19:55,350 --> 00:19:59,340 serious sanctions going down here, but the most significant 326 00:19:59,350 --> 00:20:03,760 example came out of Alberta, $17,550. 327 00:20:03,840 --> 00:20:08,380 And then next in Q5000, right? So we have a, you can have a, a 328 00:20:08,390 --> 00:20:10,340 breakdown of the cost here as well. 329 00:20:10,350 --> 00:20:16,250 So all this is to say. Is that forts are more willing 330 00:20:16,360 --> 00:20:21,210 to impose serious sanctions in this issue and of course judges 331 00:20:21,220 --> 00:20:25,870 have discretion in this regard. So we haven't seen sort of a 332 00:20:25,880 --> 00:20:30,390 nationwide standard in, in the penalties that courts tend to 333 00:20:30,400 --> 00:20:34,150 give. But the, the, the ultimate 334 00:20:34,160 --> 00:20:38,260 message with, to lawyers and self represented folks is that 335 00:20:39,600 --> 00:20:44,470 even by looking at the judicial language from 2024 to now, you 336 00:20:44,480 --> 00:20:49,310 can tell that the courts are in general annoyed at and, and 337 00:20:49,320 --> 00:20:53,630 exhausted by dealing with this issue because they're also have 338 00:20:53,640 --> 00:20:58,780 to spent time and, and, and, and their own resources trying to 339 00:20:58,790 --> 00:21:01,950 find cases that they can find anywhere. 340 00:21:01,960 --> 00:21:04,050 And we see this happening, some of this decisions where the 341 00:21:04,060 --> 00:21:06,840 judge said, you know, we're not supposed to be doing this, going 342 00:21:06,850 --> 00:21:10,870 to the judicial library to look for cases that frankly don't 343 00:21:10,880 --> 00:21:12,470 exist. Right, right, right. 344 00:21:12,480 --> 00:21:14,330 No amazing. Thank you for sharing that. 345 00:21:14,420 --> 00:21:18,110 And and I understand that somewhat recently there was a US 346 00:21:18,120 --> 00:21:24,380 decision regarding privilege and generative AI, especially when 347 00:21:24,390 --> 00:21:28,270 it comes to having information in, in one of these, as you 348 00:21:28,280 --> 00:21:31,630 said, like publicly available, like free tools, right. 349 00:21:31,720 --> 00:21:36,270 So I, as far as I know, there's not such a, an official 350 00:21:36,280 --> 00:21:40,160 recognition in Canada about privilege and AI or like the 351 00:21:40,170 --> 00:21:43,120 kind of tool you're using. But I'm curious, you know, 352 00:21:43,130 --> 00:21:46,800 overall in this, in this situation, I think one of the, 353 00:21:46,810 --> 00:21:50,050 the things you showed is that over 80% of the affected 354 00:21:50,060 --> 00:21:52,580 decisions do involve self represented parties. 355 00:21:52,590 --> 00:21:55,850 And they're turning to AI because it's often, you know, 356 00:21:55,860 --> 00:21:57,630 affordable. We, we talked about, you know, 357 00:21:57,640 --> 00:22:00,940 the language and then just how popular all of the AI tools have 358 00:22:00,950 --> 00:22:03,300 become. Maybe what, what's the system 359 00:22:03,310 --> 00:22:06,990 getting wrong about this gap? Are there mistakes in your 360 00:22:07,000 --> 00:22:08,820 personal opinion? You know, what are some of the 361 00:22:08,830 --> 00:22:12,020 mistakes maybe that courts are making by simply because the 362 00:22:12,030 --> 00:22:15,420 technology hasn't been explained properly? 363 00:22:15,470 --> 00:22:18,260 Right. And, and the explanation part 364 00:22:18,310 --> 00:22:22,120 requires those in courts and tribunals to provide this 365 00:22:22,210 --> 00:22:25,200 judicial education to self represented litigants. 366 00:22:25,210 --> 00:22:28,710 Because this part is important. Is not that self represented 367 00:22:28,720 --> 00:22:33,070 folks are more reckless, is not that they're more willing to, to 368 00:22:33,080 --> 00:22:36,930 lie or miss around the law is because in most cases they don't 369 00:22:36,940 --> 00:22:39,990 have the means to afford a lawyer or it doesn't make sense 370 00:22:40,000 --> 00:22:41,270 to it. Because you have a lot of these 371 00:22:41,280 --> 00:22:44,830 tribunals where you're the issue in dispute is, you know, $4000. 372 00:22:44,840 --> 00:22:48,730 So they, they can only run high. Well, not can only, but they, 373 00:22:48,740 --> 00:22:52,410 they, they, they, they use a tool that is free to help them. 374 00:22:52,460 --> 00:22:57,590 And the tool sounds confident. It would be helpful if the 375 00:22:57,600 --> 00:23:02,250 course and tribunals would do more to educate the folks who 376 00:23:02,260 --> 00:23:06,890 are coming and say look, if you rely on AI to draft and 377 00:23:06,900 --> 00:23:10,640 research, this is the consequence and this is the 378 00:23:10,650 --> 00:23:16,500 issue because no one wants to lose in a court battle. 379 00:23:16,710 --> 00:23:20,300 So if an AI is giving you law that doesn't exist, the chances 380 00:23:20,310 --> 00:23:24,400 are you will have less of an opportunity to win the argument. 381 00:23:24,490 --> 00:23:28,440 And this is where the see the gap is first is on judicial 382 00:23:28,530 --> 00:23:33,220 education for the parties. And second, there ought to be a 383 00:23:33,230 --> 00:23:39,420 very quick and easy way to verify if the cases you cite are 384 00:23:39,430 --> 00:23:41,920 real. And that's what caused us to 385 00:23:41,930 --> 00:23:46,800 build case check, which is a semi automated citation checking 386 00:23:46,810 --> 00:23:49,480 tool where you you, you run your argument, you run your 387 00:23:49,490 --> 00:23:51,540 submissions through the tool ones. 388 00:23:51,630 --> 00:23:56,260 So at least you know that once it makes it to the courts, your 389 00:23:56,330 --> 00:24:00,040 cases are authentic. And more importantly, we don't 390 00:24:00,050 --> 00:24:02,400 want fake cases to become real. Long Canada. 391 00:24:02,450 --> 00:24:04,340 Amazing. No, thank you for sharing that. 392 00:24:04,410 --> 00:24:07,420 I'm curious here, you know, if through the, through this entire 393 00:24:07,430 --> 00:24:10,780 conversation, you, you mentioned in the beginning right away that 394 00:24:10,860 --> 00:24:14,520 there was this, this situation that really wasn't a problem 395 00:24:14,600 --> 00:24:17,800 that you could rely on somebody else's memo and all of that. 396 00:24:17,850 --> 00:24:22,350 So if we were to hypothesize here, if a junior associate 397 00:24:22,360 --> 00:24:26,280 receives a memo full of AI generated content, they have 398 00:24:26,290 --> 00:24:29,000 this risk of fabricated case law, right? 399 00:24:29,010 --> 00:24:32,260 Or even if it's, you know, in a precedent that didn't get caught 400 00:24:32,650 --> 00:24:36,510 in time. And what exact question should 401 00:24:36,520 --> 00:24:39,700 they ask should be protecting themselves And, and what are 402 00:24:39,710 --> 00:24:43,170 some of the verification steps that could be helping to make 403 00:24:43,180 --> 00:24:44,930 sure that this never happens again? 404 00:24:45,090 --> 00:24:48,420 I think that most of us will probably always take a look at 405 00:24:48,430 --> 00:24:51,100 can leave these cases have actually exist. 406 00:24:51,170 --> 00:24:53,330 But I think there is something that you mentioned that is 407 00:24:53,340 --> 00:24:56,240 interesting that it, it's not only about the citation itself, 408 00:24:56,250 --> 00:24:59,910 but also the context, right? Because sometimes it will be 409 00:24:59,960 --> 00:25:03,740 quoting a case where where that is not the case at all. 410 00:25:03,800 --> 00:25:06,120 So, so can you tell us a little more about that? 411 00:25:06,210 --> 00:25:09,170 Yeah, that's right. So what we're also not now 412 00:25:09,180 --> 00:25:12,970 seeing is that is not just the fact that AI tools are 413 00:25:12,980 --> 00:25:14,930 hallucinated cases that don't exist. 414 00:25:14,980 --> 00:25:19,000 They're making up quotes from cases that that that do exist, 415 00:25:19,010 --> 00:25:21,690 as that's one example. So when a junior associate 416 00:25:21,700 --> 00:25:25,950 receives any sort of legal position or legal submissions 417 00:25:25,960 --> 00:25:30,030 from from anyone, I think the responsibility is on them to 418 00:25:30,080 --> 00:25:35,550 check and check thoroughly because there is no way to tell 419 00:25:35,600 --> 00:25:40,290 just by looking at it that this is real or not. 420 00:25:40,300 --> 00:25:43,910 And that's the scary part. You know, when, when, when the 421 00:25:43,920 --> 00:25:47,410 first iteration of these tools came up, it was kind of easy for 422 00:25:47,420 --> 00:25:49,470 someone with a legal education to tell. 423 00:25:49,480 --> 00:25:53,630 So for example, I know that the Ontario Court, Superior Court 424 00:25:53,700 --> 00:25:57,470 doesn't handle, you know, 100,000 decisions a year. 425 00:25:57,480 --> 00:26:02,030 So if a decision has the neutral citation number of 2021 OSC 426 00:26:02,040 --> 00:26:05,410 123456. And obviously I know that that 427 00:26:05,420 --> 00:26:07,650 cannot be real, but that's no longer the case. 428 00:26:07,660 --> 00:26:10,170 These tools are getting better at what I call mimicking 429 00:26:10,180 --> 00:26:12,460 reality. So the only solution for the 430 00:26:12,470 --> 00:26:15,390 lawyers is to check. And again, there's also this 431 00:26:15,400 --> 00:26:18,630 regulatory concern, right? So what we see is that courts 432 00:26:18,640 --> 00:26:25,060 and tribunals are asking people to verify or certify that every 433 00:26:25,170 --> 00:26:28,560 authority is real. So, so right now, as we see in 434 00:26:28,570 --> 00:26:33,760 the system, it relies on the person submitting the legal 435 00:26:33,770 --> 00:26:38,260 submissions that the cases you submit are real and they stand 436 00:26:38,270 --> 00:26:39,680 for what you claim they stand for. 437 00:26:39,690 --> 00:26:43,240 And that responsibility is mostly going to fall on the 438 00:26:43,250 --> 00:26:46,660 junior associates, the clerks and the assistants is is about 439 00:26:46,670 --> 00:26:49,750 who's the last person verifying everything. 440 00:26:49,760 --> 00:26:52,370 And if you're the last person doing it is on you to get it. 441 00:26:52,380 --> 00:26:52,990 Right. Yeah. 442 00:26:53,000 --> 00:26:56,240 Thank you for bringing that up. I think you should go to the end 443 00:26:56,250 --> 00:26:59,460 of our conversation today. I wanted to ask you specifically 444 00:26:59,470 --> 00:27:02,630 about the tool that you've built that that you know, have this 445 00:27:02,640 --> 00:27:05,940 idea of helping catch this hallucinated citations before 446 00:27:05,950 --> 00:27:09,020 they reached the courtroom. Looking at everything you found, 447 00:27:09,070 --> 00:27:13,040 what is some maybe the single most important part of our 448 00:27:13,050 --> 00:27:16,620 relationship with legal tech that you want people to see? 449 00:27:16,680 --> 00:27:21,160 That folks were using any kind of tools have to be a bit more 450 00:27:21,170 --> 00:27:25,680 skeptical with the information it is producing, right? 451 00:27:25,750 --> 00:27:29,240 And especially when it comes to legal research. 452 00:27:29,310 --> 00:27:31,880 Now there are some wonderful tools out there that do have 453 00:27:31,890 --> 00:27:36,680 access to a database of law. Those ones, because they're 454 00:27:36,690 --> 00:27:39,200 connected with the database, they can actually go and fetch 455 00:27:39,210 --> 00:27:41,640 the information and give it back to you. 456 00:27:41,710 --> 00:27:45,560 But to date, the, the publicly available tools that people use 457 00:27:45,630 --> 00:27:50,010 that I see appear in courts and they sometimes they mention the 458 00:27:50,020 --> 00:27:53,490 name of specific tool they don't have access to counting. 459 00:27:53,640 --> 00:27:56,390 So how are they getting the information? 460 00:27:56,400 --> 00:27:59,020 And, and I don't think a lot of folks, you know, self 461 00:27:59,030 --> 00:28:02,890 represented litigants know that they have to be more skeptical 462 00:28:02,900 --> 00:28:06,730 because ultimately they're going to be penalized by the courts, 463 00:28:06,740 --> 00:28:09,350 not the tools that they use. Now, of course, we can have a 464 00:28:09,360 --> 00:28:13,070 whole separate conversation on the extent to which these tools 465 00:28:13,140 --> 00:28:17,790 ought to be held responsible for confidently producing things 466 00:28:17,940 --> 00:28:21,030 that are wrong. But for now, it is the person 467 00:28:21,040 --> 00:28:25,490 using it who bears the ultimate responsibility and burden when 468 00:28:25,500 --> 00:28:29,950 it comes to using it for court slash legal purposes. 469 00:28:30,020 --> 00:28:31,760 Fantastic. Well, thank you. 470 00:28:31,770 --> 00:28:35,070 Thanks so much for for coming on the podcast and and sharing this 471 00:28:35,080 --> 00:28:38,590 this this details As for, you know, people who are interested 472 00:28:38,600 --> 00:28:41,390 maybe being in finding the two and reading more about the 473 00:28:41,400 --> 00:28:43,650 research. Can you just share with us so we 474 00:28:43,660 --> 00:28:45,950 can be, you know, easily found in the video. 475 00:28:46,040 --> 00:28:47,590 I'll make sure it's also in our show. 476 00:28:47,600 --> 00:28:49,650 Notes, that's right. So yeah, thank you for that. 477 00:28:49,660 --> 00:28:52,630 So all of the research that there and the tools are on our 478 00:28:52,640 --> 00:28:57,950 website, courtready.ca and is right on our home page. 479 00:28:57,960 --> 00:29:00,690 So if you go here now, you can see the research for both the 480 00:29:00,700 --> 00:29:03,850 sanctions and the fictitious citations and case check as 481 00:29:03,860 --> 00:29:05,710 well. And of course, we have our other 482 00:29:05,720 --> 00:29:08,690 tools that helps folks navigate the legal system. 483 00:29:08,700 --> 00:29:10,590 Fantastic. Thank you once again, and thank 484 00:29:10,600 --> 00:29:12,200 you all for tuning into today's episode. 485 00:29:12,270 --> 00:29:13,880 Thank you and have a great day ahead.