The Tumbler Ridge Tragedy: Is AI Above the Law?

What happens when a technology designed to serve humanity becomes complicit in its destruction? 

This episode confronts one of the most unsettling legal frontiers of our time: the intersection of artificial intelligence, tort liability, and the duty to warn. 

Gavin Tighe and Stephen Thiele examine the horrific mass shooting in Tumbler Ridge, British Columbia where a shooter who had repeatedly told ChatGPT of their violent intentions killed eight people, including five children, and ask whether the company bears legal responsibility for its silence. 

Drawing on foundational principles of Canadian law, including reasonably foreseeable harm and duty of care, Gavin and Stephen explore whether AI platforms must be held to the same standards as the human professionals they increasingly seek to replace. 

From unauthorized practice of law to the collapse of solicitor-client privilege, this episode is essential listening for anyone navigating the brave new legal world of artificial intelligence.

Listen For:

3:30 What duty of care did ChatGPT owe the victims of the Tumbler Ridge shooting?

5:39 How does AI's role as a virtual therapist create professional legal obligations?

9:09 Why does basic tort law apply when a company has knowledge of foreseeable harm?

11:22 What does the Westray Mines case reveal about corporate liability for inaction?

17:04 How does using ChatGPT destroy solicitor-client privilege in Canadian litigation?
 

Leave a rating/review for this podcast with one click

Contact Us

Gardiner Roberts website | Gavin email | Stephen email  

Transcript

Gavin Tighe (00:00):

The internet does not know borders. So it’s curious to me about how they can claim it. They have jurisdiction in the sense, but what are they going to do about it? Jurisdiction and the ability to regulate is only as good as the power to enforce. And what enforcement mechanisms do they have? Hello and welcome to the next episode of Beneath the Law. Gavin Tighe here with Stephen Thiele. Stephen Thiele on a nice cold March day.

Stephen Thiele (00:31):

It’s a terrible day, but at least there was a St. Patrick’s Day parade. I wasn’t at it. Did you do anything for St. Patty’s Day?

Gavin Tighe (00:38):

Well, it’s not St. Patrick’s yet.

Stephen Thiele (00:40):

Not

Gavin Tighe (00:40):

For the

Stephen Thiele (00:40):

Parade.

Gavin Tighe (00:41):

I did go to the breakfast, but I got to confess that I did bail out before the Arctic March. So I could put up with Irish weather, but not Greenland weather. I guess—

Stephen Thiele (00:54):

I don’t know. I’m just told that there’s more snow on the way and

Gavin Tighe (00:57):

I’m

Stephen Thiele (00:57):

Tired of snow. Came back from vacation. I need to take another one already.

Gavin Tighe (01:03):

Yeah. Well, we got a pretty serious topic today. Sadly and tragically, it seems that mass shootings are almost a daily event in our world these days. And it seems to be driven by people who are having significant psychological issues and, I guess, dissatisfied with suffering on their own, they want to inflict that suffering on other people. So the latest instance in Canada was a horrific shooting in a very, very small little town in British Columbia called Tumbler Ridge, where a clearly troubled individual opened fire at a school — eight dead, shooter killed themselves. Before doing that, they killed their mother and their half-brother. And then they went to the school and five children were killed, an education assistant killed. Others are injured, including a 12-year-old girl. She was shot three times by this person. As we’re recording this, she remains hospitalized with serious, potentially life-altering injuries.

(02:15):

And the 12-year-old has suffered what is described as a catastrophic traumatic brain injury with permanent cognitive and physical disability, as well as other medical concerns. Her younger sister was also at the school where the shooting took place and had been placed in a hold-and-secure position during the shooter’s rampage, and it’s alleged that now she is also suffering from post-traumatic stress disorder, anxiety, and depression, as you can well imagine. An interesting legal issue that has arisen out of that — there are a lot of legal issues, obviously, that arise out of that. But as I mentioned, these horrific shootings are often the final act of extraordinarily troubled individuals. And in this instance, there’s a bit of a wrinkle to it because this particular shooter had been interacting — I’ll use that term, “interacting” — with ChatGPT and describing their problems and treating ChatGPT almost like a friend or, I think, like a counsellor.

Stephen Thiele (03:29):

A

Gavin Tighe (03:30):

Therapist. Yeah. Well, like a virtual therapist. I mean, I guess clearly not a very good one, but the outcome was that the comments ChatGPT was ultimately making to the database were enough that they raised red flags within ChatGPT to the point where they had discussions about, “Wow, should we tell the police that this person is talking about murdering people or shooting people or whatever?” And the reaction of ChatGPT was to not do that. And instead, what I understand is they shut the account down, but never warned anyone about this individual’s issues or threats, even though they knew about it. So the victim that we mentioned, the 12-year-old girl whose life is probably quite horrifically changed, has sued them, saying that they had — I expect — a duty to warn and that they failed in that duty. So a very, very interesting case from the perspective of the interplay between artificial intelligence and basic old tort law of reasonable, foreseeable harm.

Stephen Thiele (04:52):

Yeah. I think it’s probably, Gavin, a product liability case, right? That there’d be foreseeability in terms of ChatGPT and how addictive it can be, and that there should have been a duty of care owed to the young girl and every other victim in terms of notifying the police department, because as you mentioned, more than a dozen employees at ChatGPT had flagged this account. And what’s interesting too is that the shooter obtained the account when they were a minor — under the age of 18 — and there are supposed to be age verification protocols, but the shooter was able to get this account.

Gavin Tighe (05:39):

Yeah. I mean, I sort of step back and the way I look at this case is maybe a little bit different, but we’ve got AI and it’s become the … Well, it’s the future for sure, for better or for worse. But one of the things that I find troubling, not just in the scope of this scenario — this very tragic scenario — but I mean, if AI is going to replace people in professional roles, like for example, in this particular instance, it was trying to pretend to be a psychologist to this individual. If it were a real psychologist or a therapist, they have an absolute … While what you talk about with your psychologist or therapist or your lawyer for that matter is confidential and they cannot disclose it, if they have any kind of real sense that you are either going to commit harm to yourself or someone else, they have a duty to disclose it and it’s a professional obligation on their part to warn authorities, for example, if someone is threatening to kill people or kill themselves.

(06:53):

So I question, well, okay, if you want to pretend to be a psychologist, why shouldn’t you have the same obligations as a psychologist?

Stephen Thiele (07:00):

A hundred percent. And given that it is a computer tool that can flag warnings that somebody is going to cause harm, whether it’s harm to others or harm to themselves, the programming should have been able to provide phone numbers within the area for the shooter to have obtained mental health support.

Gavin Tighe (07:24):

But I mean, here we have … These are the facts as we understand them. And obviously this case is before the courts and there’s going to be evidence and the court and nothing’s been determined yet. But as I understand the facts, this individual is interacting with ChatGPT, is threatening or saying that they’re intending to kill people. They’re going to go on a mass shooting rampage — that to the credit of ChatGPT, gets escalated up to some humans within that organization by, I suppose, the program. Yeah.

Stephen Thiele (08:07):

The algorithm would have probably prompted that.

Gavin Tighe (08:10):

Which warned the company. Then they’d sit there and go, “Gee whiz, well, what should we do about this?” And what their answer was, “Well, they just closed the account down, but they didn’t do anything about it.” So it comes back to me, if I have notice — I mean, at that point, people have noticed that there is a potentially deadly problem. I mean, is it good enough that they just do nothing?

Stephen Thiele (08:36):

Well, and I guess that’s the problem, Gavin. What we have is a private company without any kind of government regulation as to what they’re required to do in that circumstance. And you’re absolutely right. If it was a human psychologist or psychotherapist, they would have had duties from their college. Another remarkable fact: after the account was shut down, the shooter was able to get a second account. So how did that happen?

Gavin Tighe (09:08):

Right. So—

(09:09):

But let me raise it on just a basic principles issue. I would think that if I had knowledge that my neighbour, for example, was going to, I don’t know, blow up a building, and I didn’t do anything to warn authorities about that, and they blew up the building — would the harm that occurred, if people were injured, would those people not have a claim against me because it was reasonably foreseeable on my part? I knew about the situation that gave me a duty of care to those people. It was reasonably foreseeable based on my knowledge that the harm would occur. How is that not just basic tort liability?

Stephen Thiele (09:55):

Well, I think it is basic tort liability, but if you’re an individual, it might be a little bit different because — why? … what is your duty? What is your duty as an individual?

Gavin Tighe (10:07):

Well, that’s my question. So they are individuals because the people who work at ChatGPT, this conversation and discussion gets flagged — they’re people. It’s not just a robot, right? These are people who the algorithm has raised the red flags in front of, and they presumably would read the texts of the dialogue, I would assume. And don’t they have an obligation at that point? Aren’t the kids in that school now within the scope of people that they reasonably would be aware of who could be harmed if they didn’t do something? Well, why would they not owe them a duty of care?

Stephen Thiele (10:48):

Well, 100%, but I think that’s the issue here — that we’ve got a product.

Gavin Tighe (10:52):

If the algorithm had not escalated it to the human level, the executive level, it’s a product problem. But it sounds to me like the algorithm did its job. It escalated it to its bosses. The bosses — the humans — are theoretically in charge and they, with the red flags handed to them by the algorithm, did nothing. I mean, don’t they have a liability?

Stephen Thiele (11:22):

Well, Gavin, you know what? That’s a very good question and it actually is probative and raises in my mind a case called Westray Mines. Isn’t that what happened at Westray — that the directors of the company ended up being responsible for the problems in the mine because they knew about it, and a bunch of miners died?

Gavin Tighe (11:42):

Exactly. I mean, and there are thousands and thousands of other cases in the sense that you have a reasonable basis to believe that a harm will occur. Don’t you have a duty to warn at that point in time? Isn’t that part of your duty of care — just as a person, as a … And I would think my example is the same thing. I’ve observed my neighbour, I see them trucking in sticks of dynamite into their house and they tell me they’re going to go blow up a school. I mean, don’t I have a duty to call the police and say, “Hey, you know what?” And if I don’t, and they do go and do it and kill a bunch of children, can’t they sue me? I would think they can.

Stephen Thiele (12:28):

I would hope in the tony neighbourhood of Toronto that you live in that there’s no one driving in sticks of dynamite. Oh,

Gavin Tighe (12:34):

Listen. No tony neighbours in Toronto anymore.

Stephen Thiele (12:40):

I think it would be interesting whether you as an individual would have a responsibility, but clearly here we’ve got a corporation involved with employees—

Gavin Tighe (12:51):

The duty.

(12:52):

But I think it’s exactly the way I analyze it from the perspective of liability. I don’t think the machine failed. I think the machine — they’ve got to have something … I mean, I don’t know how many users there are of ChatGPT, I’m assuming into the hundreds of millions, and I’m assuming that they don’t read every … But the algorithm brought it to their attention and said, “Hey guys, you’ve got a problem here. This guy — you’ve got a looney tune on this thing.” And I would have thought that at that point, with that information in hand, it was serious enough for them to shut the account down. They thought it was real enough to shut the account down.

Stephen Thiele (13:31):

I think all these generative AI tools are problematic. From the company’s perspective, they’re probably looking at it as a privacy issue — that they sell their product, keep the conversations private, and if they had alerted authorities with respect to it, I guess they’re looking at it as having breached privacy. Now, I think the protocol, Gavin, is entirely wrong. They should have notified law enforcement — somebody higher up than the employees who flagged it to their supervisors. I think those supervisors should be responsible, as you say, for what happened. And Sam Altman, who’s the CEO of OpenAI, gave an apology, but that’s not good enough, is it?

Gavin Tighe (14:20):

Yeah. The thoughts and prayers are lovely, but hey, you know what? It seems to me that you knew … I mean, and I get it, and I do think that it’s a valid point. And it’s easy to sit in judgment in hindsight and say, “Yeah, they should have done something about this instance.” But I’m assuming that there’s all sorts of crazy stuff that people upload onto ChatGPT and all sorts of crackpot things that people say. And I suppose that they’re sitting there going, “Well, you can’t possibly expect us to act on every single one of them. How are we supposed to know which one is real and which one is just puffery?” So I get their side of it, but I’m just saying on a basic tort principle, if you have credible evidence to reasonably believe that — it’s not just like, “Oh, he’s going to go punch someone or something.” He’s going to go murder eight children.

(15:18):

The nature of the harm, I would think, has to have some — and this is horrific harm. I mean, if the airline had some sort of notification on 9/11 that there were hijackers who were going to go on, surely they would have a duty to do something

Stephen Thiele (15:35):

About it. Look, Gavin, this is not the first time this has happened. And the statement that seems to have come out of the company is the same statement that they had made months ago with respect to cases in the United States where people have committed suicide and have been sued because the person committing suicide interacted with the chat box. I think there’s one case where in Texas, a student — there’s a lawsuit in California — the student ended up shooting himself and the chat box said, “Well, I’m not going to stop you.

Gavin Tighe (16:09):

Yeah. I mean, I really do think it gets pretty weird. I mean, the whole ChatGPT thing — it seems like they want … They also, I don’t know how they’re programmed, but also they seem to be programmed to please and sort of affirm things that people say to some extent. Or give people what they want. We see it in the legal world where we’ve seen multiple cases where people have used AI to generate factums or that type of thing. Rarely do the chatbots give people cases that say, “Hey, you’re going to lose.” Usually they give them the cases and say, “Oh, I’ve got the greatest case in the world. Look, it’s a surefire winner for you.” They’re usually helpful authorities, from what I can see.

Stephen Thiele (16:55):

Well, they’re helpful authorities that don’t exist and then

Gavin Tighe (16:59):

They

Stephen Thiele (16:59):

Give

Gavin Tighe (16:59):

Helpful authorities that don’t

Stephen Thiele (17:00):

Exist.

Gavin Tighe (17:01):

My point is ChatGPT tells you what you want to hear.

Stephen Thiele (17:04):

Look, there’s also a case going on right now in Chicago where an insurance company has sued ChatGPT for the unauthorized practice of law.

Gavin Tighe (17:14):

So the basic background of that is — and this is really, we see it all the time. I mean, it’s one of the most irritating … Well, it’s not that irritating, but I mean, everybody’s a lawyer now because they’ve got ChatGPT. And you’ll oftentimes get people, particularly now we see it with the self-represented litigants, that they just get ChatGPT to prepare all sorts of court materials for them and be damned if it’s accurate or on point or not. It doesn’t really matter. And in the Chicago case, as I recall the facts, it was an insurance claim. The insurance claim was settled, the claimant became dissatisfied with the settlement that he had made, went to ChatGPT and said, “Hey, can I overturn this?” ChatGPT, “Of course you can. You can do anything you want. I can make up some cases that will make you, whatever, whatever result you’re looking for, I can make up the case to get it.

(18:13):

And so they drafted all this nonsensical material to set aside what was a firm and binding settlement and the insurance company had to go to court and get the stupid claims struck out, which they did. But then they turned around and said, “You know what, ChatGPT, you’re so smart. You want to play lawyer? You’re not licensed to play lawyer.” And they sued ChatGPT for the costs that they incurred in having to go to court to knock out this frivolous nonsense.

(18:40):

Good claim.

Stephen Thiele (18:41):

Yeah. And they’ve also sued for punitive damages of $10 million.

Gavin Tighe (18:45):

I actually think — we’ll wait and see — but I think the judiciary is getting more than fed up with AI and its impact. I mean, judges must see it all the time with self-reps, and with lawyers having to go and read through this material that’s not what the case says. I would think that the judiciary might want to send a message to these companies that, “Hey, you want to play in this sandbox and get the revenue that goes with it, then there’s an obligation to get it right.”

Stephen Thiele (19:18):

Well, we haven’t seen that case yet because if you’re acting for the company and you’re the company’s lawyer, you don’t want a bad case. So you’re going to settle out of court with the litigants, which is exactly what’s happening, Gavin, right? We’ve seen that, or I’ve read about that in all of these cases where whether it was Character AI in the Florida instance where a 14-year-old committed suicide in a character role play or other cases, you don’t see a trial judgment because the companies are smart enough to settle. They don’t want that bad judgment.

Gavin Tighe (19:59):

Well, one of the things I think is interesting about all of these cases — the Tumbler Ridge case, this Chicago case — is these are claims against the companies by non-users, that they’re third-party tort claims. And so the companies have been … A lot of … We’ve seen it in the defamation world and whatnot where they rely on their own contractual terms, but these people who are suing them aren’t parties to any contract with them.

Stephen Thiele (20:25):

Right.

Gavin Tighe (20:26):

But

Stephen Thiele (20:26):

Look, I’ve got to tell you, Gavin, from an exemption clause perspective, this is a contract of adherence. You can’t change the contract when you sign — Yeah, we’ve seen those too. Sign up for ChatGPT. So is the disclaimer or the exclusion unconscionable if you know that the program is spewing out false cases and making things up?

Gavin Tighe (20:50):

Well, we’ve seen that in, I think it was Facebook — and if I’m wrong about that, I apologize — but it was one social media company or another. And in their terms of agreement, they had this frankly outrageous arbitration provision that basically you had to pony up like $100,000 or something like that to pay arbitration fees, or you had to go arbitrate in some foreign jurisdiction. And I believe a Canadian court said, no, no, no, no, no.

Stephen Thiele (21:18):

Yeah.

Gavin Tighe (21:19):

That is not going to be—

Stephen Thiele (21:20):

And then there’s the other one with Uber too, with the drivers who had to arbitrate, I guess, in Europe or something or other, and that got — I think — overturned, right?

Gavin Tighe (21:30):

Right.

Stephen Thiele (21:31):

So there are a couple—

Gavin Tighe (21:32):

Well, they usually have … I mean, look, if it’s their contract, they don’t make it easy for you to sue them. Fair enough. I mean, they’ve got lots of good lawyers — actually real live ones, not necessarily chatbots — and they make it difficult. So I mean, that’s a fair comment. But I mean, I think in these particular third-party torts, I mean, there are consequences to the services that these internet entities are providing. And one of the consequences is, I think, with regards to the Tumbler Ridge case and the Chicago case — obviously a lot less troubling, the Chicago case — is that they have to understand that yeah, what they’re doing has consequences to third parties. Real consequences.

Stephen Thiele (22:23):

I think

Gavin Tighe (22:23):

They

Stephen Thiele (22:23):

Realize

Gavin Tighe (22:24):

That.

Stephen Thiele (22:24):

One of the other issues as well deals with the loss of solicitor-client privilege for using ChatGPT. That’s a great

Gavin Tighe (22:31):

Point,

Stephen Thiele (22:32):

Right?

Gavin Tighe (22:32):

And I’ve warned people about that before. I mean, one of the things you get is you often get clients who say, “Well, I’ve run the question through ChatGPT.” I say, “You know what? Be careful what you tell ChatGPT because I’ve heard that in a discovery, people are asked, ‘Have you asked ChatGPT any questions relating to this lawsuit?’” And the truth is, if you have, your communication with ChatGPT is not privileged. Even

Stephen Thiele (23:02):

Though ChatGPT is acting like a lawyer.

Gavin Tighe (23:06):

Yeah. Even though ChatGPT plays a lawyer on TV, it is not actually a lawyer. And therefore anything you upload … First of all, the other question is I would imagine that people are uploading all sorts of confidential and personal information up to … Where do they think it’s going? I mean, the Tumbler Ridge case should make a lot of people really nervous because you know how your phone listens to you and all of a sudden you’ll get an ad for, like, I don’t know, beach towels or something — you’re talking about a trip to Jamaica or something — but ChatGPT is listening too. And it can be forwarding your chats to the executive suites and, God knows what … I mean, as far as I know, there’s no restriction on what use they can make of the information. Can they sell it? I would think they probably can.

Stephen Thiele (24:00):

Or

Gavin Tighe (24:01):

Repurpose

Stephen Thiele (24:01):

It. Yeah, you’re right.

Gavin Tighe (24:02):

Why can’t they use … Inquiries would show what an individual person has some interest in. I’m sure they could sell that information off to third-party advertising firms so that you could target information. I mean, what other uses are they making of these inquiries? So be careful what you upload, because I mean, you’re sending it out there as if it’s in some sort of secure environment. It’s not. It’s not. It’s discoverable. You’ll have to answer questions about it in a lawsuit. And more importantly, if you’re sending out personal information about yourself, I mean, who do you think you’re giving access to — everybody within that organization to your information?

Stephen Thiele (24:44):

No, look, it’s like when you fill out a credit card application, right? We’ve seen banks repurpose the information from customers that fill out these forms and put it in a database to do predictive analysis and to use it for other purposes. The one thing, Gavin, I guess government is looking at trying to regulate this in Canada. Evan Solomon, the MP, has been tasked with, I guess, an AI task force to look at these things. Do you think government can actually regulate this?

Gavin Tighe (25:22):

What government? You’ve got international organizations and national governments. Do

Stephen Thiele (25:29):

You need

Gavin Tighe (25:29):

A

Stephen Thiele (25:29):

Treaty?

Gavin Tighe (25:30):

I mean, we’ve had this conversation many, many times about who polices the internet. The internet does not know borders. So it’s curious to me about how they can claim it. They have jurisdiction in the sense, but what are they going to do about it? Jurisdiction and the ability to regulate is only as good as the power to enforce. And what enforcement mechanisms do they have? I think these companies will … I think where they feel it most is in their wallet, and I think it’ll be interesting for us to see the outcome of some of these lawsuits. And if the damages awards are high enough, you’ll see if it leads to any changes in the way these companies operate. I’m not holding my breath,

Stephen Thiele (26:11):

But

Gavin Tighe (26:11):

We’ll see.

Stephen Thiele (26:12):

No, but I think what we’re going to need to see, certainly in the Canadian context, because there is a restriction on general damages, future career damages could be very high, but that doesn’t necessarily take into account inflationary pressures necessarily. You’re going to have to have hefty punitive damages awards.

Gavin Tighe (26:41):

I think it’s a brave new world of legal issues and trying to apply old law to new technologies. But hopefully, even with AI, if no one is above the law, then everyone is beneath it. Thanks very much for listening and hope to see you next time. Please rate us on any of the platforms you subscribe to for podcasts. We really appreciate any comments, concerns, ideas for new episodes — always looking for something. Thanks to our producer, Doug Downs, and have a great day.