"In an argument that appeared to flabbergast a small claims adjudicator in British Columbia, the airline attempted to distance itself from its own chatbot's bad advice by claiming the online tool was "a separate legal entity that is responsible for its own actions."
"This is a remarkable submission," Civil Resolution Tribunal (CRT) member Christopher Rivers wrote.""
From https://www.cbc.ca/news/canada/british-columbia/air-canada-c...
IANAL, but it's astounding they took that as their defense, rather than pointing to a line (I hope?) in their ToS that says "This agreement is the complete terms of service, and cannot be amended or changed by any agent or representative of the company except by ... (some very specific process the bot can't follow)". I've seen this mentioned in several ToSs, I expect it to be standard boilerplate at this point ...
That does make sense, but on the flipside, let's say that they start advertising discounts on TV, but when people try to pay the reduced rate they say "according to our ToS that TV ad was not authorized to lower the price".
Obviously that wouldn't fly. So why would it fly with the AI chatbot's advertising discounts?
You'd normally expect a TV ad to be authorized to make offers.
You wouldn't normally expect an AI chatbot to be authorized to make offers. Its purpose is to try to answer common questions and it has been widely covered in popular media that they hallucinate etc.
I disagree. I expect any credible offer a company makes in an advertisement, on its website, using a chatbot, or through a customer service agent to be authorized by the company. Surely a corporation with lots of resources knows better than to program a chatbot to make fake offers; they'd get sued.
And they did get sued. Next time maybe they'll make sure software they connect to their website is more reliable.
They didn't program it to do that, it's a characteristic of the technology that it makes mistakes. Which is fine as the public learns not to blindly trust its answers. It seems silly to assume that people won't be able to figure that out. People are capable of learning how new things work.
This is like the people who set the cruise control in their car when it first came out and then climbed into the back of the car to take a nap. That's not how it works and the technology isn't in a state where anybody knows how to do better.
I agree with your cruise control analogy in a sense, but I think it's Air Canada that's misusing the technology, not the customer. If they try to replace customer service agents with chatbots that lie, they need to be prepared to pay for the results. I'm glad they're not allowed to use such unreliable, experimental technologies in their airplanes (737 Max notwithstanding).
There's absolutely a technology available to make a chatbot that won't tell lies: connect a simple text classifier to a human-curated knowledge base.
The result would be a de facto ban on AI chatbots, because nobody knows how to get them not to make stuff up.
If you use unreliable technology in an airplane, it falls out of the sky and everybody dies. If you use it in a chatbot, the customer can e.g. go to the company's website to apply for the discount it said exists and discover that it isn't there, and then be mildly frustrated in the way that customers commonly are when a company's technology is imperfect. It's not the same thing.
But then it can only answer questions in the knowledge base, and customers might prefer an answer which is right 75% of the time and can be verified either way in five minutes than to have to wait on hold to talk to a human being because the less capable chatbot couldn't answer their question and the more capable one was effectively banned by the government's liability rules.
No, the result would be a de facto ban on using them as a replacement for customer service agents. I support that for the time being since AI chatbots can't actually do that job yet because we don't know how to keep them from lying.
They could put a disclaimer on it of course. To be sufficiently truthful, the disclaimer would need to be front and center and say something like "The chat bot lies sometimes. It is not authorized to make any commitments on behalf of the company no matter what it says. Always double-check anything it tells you."
But what does that even mean? If Ford trains a chatbot to answer questions about cars purely for entertainment purposes, or to get people excited about cars, a customer could still use it for "customer service" just by asking it questions about their car, which it might very well be able to answer. But it would also be capable of making up warranty terms etc., so you've just banned that thing and anything like it.
It's pretty unlikely we could ever keep them from lying. We can't even get humans to do that. The best you could do is keep them on a script, which is the exact thing that makes people hate existing human customer service reps who can't help them because it isn't in the script.
Which is exactly what's about to start happening, if that actually works. But that's as pointless as cookie banners and "this product is known to the State of California to cause cancer".
It's all in how it's presented and should not be up to the customer or end-user to understand how technology running on the company's server, which might be changed at any time might behave unreliably.
I expect something that's presented as customer service not to lie to me about the rebate policy. As long as what it says is plausible, I expect the company to be prepared to cover the cost of any mistakes, especially if the airline only discovers the mistake after I've paid them and taken a flight. Compensating customers for certain types of errors is a normal cost of doing business for airlines, and the $800 CAD this incident cost the airline is not an exorbitant amount. The safety valve here is that judges and juries do test against whether a reasonable person would believe a stated offer or policy; I can't trick a chatbot into offering me a billion dollars for nothing and get a court to hold a company to it.
If Ford presents a chatbot as entertainment and makes it really clear at the start of a session that it doesn't guarantee the factual accuracy of responses, there's no problem. If they present it as informational and don't make a statement like that, or hide it in fine print, then it says something like "the 2024 Mustang Ecoboost has more horsepower than the Chevrolet Corvette and burns less gas than the Toyota Prius", they should be on the hook for false advertising to the customer and unfair competition against Chevrolet and Toyota.
Similarly, if Bing or Google presents a chatbot as an alternative to their search engine for finding information on the internet, and it says "Zak's photography website is full of CSAM", I'm going to sue them for libel.
Sure, but a billion people could each trick it into offering them $100, which would bankrupt the airline.
But all you're really doing is requiring everyone to put a banner on everything that says "for entertainment purposes only". Because if something like that gets them out of the liability then that's what everybody is going to do. And if it doesn't then you're effectively banning the technology, because "have it not make stuff up" isn't a thing they know how to do.
Courts probably aren't going to enforce any promise of money for nothing or responses prompted by obvious trickery, but they might enforce promises of discounts, and are very likely to enforce promises of rebates as the court in this case did.
If that means companies can't use chatbots to replace customer service agents yet, so be it.
And if I saw that disclaimer, I wouldn't use the tool. What's the point if you can't trust what it says. Just let me talk to a human that can solve my issue.
That's the point of it -- you don't have to wait on hold for a human to get your answer, and you could plausibly both receive it and validate it yourself sooner than you could get through to a human.
I think that banning lying to customers is fine.
ChatGPT is presumably capable of making something up about ChatGPT pricing. It should be banned?
If, instead of a chatbot, this was about incompetent support reps that lied constantly, would you make the same argument? "We can't hire dirt-cheap low-quality labor because as company representatives we have to do what they say we'll do. It's so unfair"
It isn't supposed to be a company representative, it's supposed to be a chatbot.
If Microsoft puts ChatGPT on Twitter so people could try it, and everybody knows that it's ChatGPT, and then it started offering companies free Windows licenses, why should they have to honor that? It's obvious why it might do that but the purpose of letting people use it wasn't so it could authorize anything.
If the company holds a conference where it allows third party conference speakers to give talks, which everybody knows are third parties and not company employees, should the guest speakers be able to speak for the company? Why would that accomplish anything other than the elimination of guest speakers?
It sounds like you meant to say that they didn’t _intentionally_ program it to do that. They didn’t find the system under a rock and unleash it on the world; they made it.
Most of these companies didn't make it, they took an existing one and fed it some additional information about their company.
So what?
Yes. And therefore people should be able to assume that the answers are correct.
Some people have heard of ChatGPT, and some of those have heard that they hallucinate, sure. But that's still not that many people. And they don't know that a question answering chat bot like this is the same technology!
Why is that a necessary requirement? Something can be useful without it being perfect.
If I am trying to interact with a company and they tell me to use their chatbot, I expect that chatbot to provide me with accurate answers 100% of the time (or to say it can't help me in the event that I ask a question that it's not meant to solve, and connect me to a representative who can).
If I have to double-triple check elsewhere to make sure that the chatbot is correct, then what's the point of using the chat bot in the first place? If you can't trust it 99% of the time, or if the company says "use this, but nothing it says should be taken as fact", then why would i waste my time?
This is why I’m a bit vexed by all the hype around LLMs. It reminds me of talking to a friend’s mother who was suffering from dementia - she could have a perfectly lucid conversation with you and then segue into stories that were obviously fictions that existed only within her head. She was a nice lady, but not someone who you would hire to represent your company; she was considered disabled.
Awhile back another commenter called them a “demented Clippy” which about sums them up for me.
Yeah totally. LLMs have a lot of awesome use cases. But as chatbots, they need a lot of guardrails, and even then, I'm highly skeptical if they improve the experience over a simple searchable FAQs or docs.
Because you can ask it a question in natural language and it will give you an answer you can type into a search engine to see if it's real. Before you didn't know the name of the thing you were looking for, now you do.
The rate at which it makes stuff up isn't 99%, is the point. For common questions, better than half of the answers have some basis in reality.
I would expect that if I was talking to an official tool that the company provides to interact with to be authorized to give me information (including discounts and offers) to be accurate and true.
What would it take to disabuse you of that notion now that your expectations have been observed to be in conflict with reality?
What you're describing isn't what you expect, it's what you wish were the case even though you know it isn't.
Of course language models can't be trusted, but it's not the customer's problem to think about chatbot's purpose, how it's implemented and whether it hallucinates or not.
If it was approved by the company, yes. But you wouldn't want Braniff Airlines to put out an ad for SouthWest Airlines advertising rock bottom lifetime tickets and have those be valid...
I think only software engineers would think this. I don't think it is obvious to a layperson who probably has maybe never even used one before.
How do those clauses actually work? If a rep does something nice for you (like give you something for free), could the airline say it never agreed to that in writing or whatever and demand it back? How are you supposed to know if a rep has authority to enter into an agreement with you over random matters?
But, to your question, my guess is that would basically be telling people not to avoid their chatbot, which they don't want to do.
It's more to shield them from cases like a rep gifting you free flights for life.
I realize the intention but I'm wondering how it works legally given what the terms actually say.
What you are or aren't entitled to is written down in the terms of service. Support agents can only help interpret the terms for you. They may be authorized to go beyond that to some degree, but the company will also have the right to reverse the decision made by the agent.
Those ToS statements overreach their capabilities a lot of the time. They're ammunition against the customer, but don't always hold up in the legal system.
Beyond the chatbot's error and the legal approach they took, this bad PR could have been avoided by any manager in the chain doing the right thing by overriding things and just giving him the bereavement fare (and then fixing the bot/updating the policy).
I guess the original issue pointed by the judge would still stand: how am I supposed to know which terms are to be assumed true and valid? Why would I assume a ToS hidden somewhere (Is it still valid? does it apply to my case? Is it relevant and binding to my jurisdiction?) is to be considered more trustworthy than an Air Canada agent?
Courts often rule that you can’t use ToS to overcome common sense. ToS are not a get out of jail free card if your company just does stupid things.
How is that enforceable? In many cases this is carte blanche for company representatives to lie to you. No one is going to read the ToS and cross reference it with what they're being told in real time. Moreover, if a customer was familiar with the ToS they would not be asking questions like this of a chatbot. The entire idea of having a clause like this while also running a "help" chatbot that can contradict it seems like bad faith dealing.
How is this different from me getting one of my friends to work at Air Canada and promise me a billion dollars to cancel my flight?
Will Air Canada be legal for my friend going against company policy?
That's fraud because you're in cahoots with your friend.
If a random AC employee gave you a free flight, on the other hand, you'd be entitled to it.
Anyway, the chat bot has no agency except that given to it by AC; unlike a human employee, therefore, its actions are 100% AC actions.
I don't see how this is controversial? Why do people think that laws no longer apply when fancy high-tech pixie dust is sprinkled?
A random AC employee drinks too much and says "You are entitled to free flights for the rest of your life." Is Air Canada liable?
Since when are contracts enforceable when one party is drunk?
A random AC employee who is having a bad day and hates his employer says "You are entitled to free flights for the rest of your life." Is Air Canada liable?
Valid contracts usually require consideration
No, because no reasonable person would think that they had the authority to authorize that. Remember, the legal system is not computer code - judges look at things like intent and plausibility.
No, because that's not "reasonable". My dad jokes that he's made a career off of determining what is "reasonable" and what isn't, and he's a contract attorney.
If you were standing at the customer service desk, and instead they said: "sorry about the delay, your next two flights are free", then all of a sudden this is "reasonable".
The company would be entirely within their rights to say 'this employee was wrong, that is not our policy, goodbye!'. This happens all the time with more minor incidents.
No idea about the US but this very same case was tested in France and some part of Germany in the late 90s when some PayTVs (Sky or Canal+, can't remember) tried to cancel multiple subscriptions offered with an extremely aggressive pricing by some of their agents. Courts concluded that the signed agreements superseded the official pricing and they had to offer the service for the entire length of the original subscription.
The difference is that was a signed agreement.
This chatbot merely said something was possible, no legally binding agreement occured.
Where I live, "meeting of the minds" is necessary for a contract. Written or not. In this case, that meeting didn't happen. Due to the bullshit generator employed by Air Canada.
So there was no contract but a consumed flight. The court has to retroactively figure out a reasonable contract in such cases. That Air Canada couldn't just apply the reduced rate once they learned of their wrong communication marks them as incredibly petty.
That's far less likely to be true if the customer buys something based on the employee's erroneous statement. I suspect in an otherwise-identical case with a human customer service agent, the same judge would have ruled Air Canada must honor the offer.
And if a random AC employee said[0] they'd give you a billion dollars, you wouldn't be entitled to it because any judge or jury hearing the case would say a reasonable person wouldn't believe that. Unlike computers, which do exactly what they're told[1], the legal system applies sanity checks and social context.
[0] perhaps because they're disgruntled and trying to hurt their employer
[1] generative models are not an exception; they're a way of telling computers to generate text that sometimes contains falsehoods
I'm sure that if the bot had said that the airline would raise your dead relative from the grave and make you king of the sky or something equally unbelievable the courts wouldn't have insisted Air Canada cast a crown and learn necromancy.
And if it was a billion dollars?
Because their source of income depends on sprinkling fancy high-tech pixie dust!
The computer only does what they told it to.
What they told it to do was to behave very unpredictably. They shouldn’t have done that.
Not these ones...
These ones do what they "learned" from a lot of input data using a process that is us mimicking how we think brains could maybe function (kinda/sort off with a few unbiological "improvements").
Yes, these ones. Somebody told the computer to do all those things you just wrote.
Maybe this makes the point better:
Say your webserver isn't scaling to more than 500 concurrent users. When you add more load, connections start dropping.
Is it because someone programmed a max_number_of_concurrent_users variable and a throttleExtraAboveThresholdRequests() function?
No.
Yes, humans built the entire stack of the system. Yes every part of it was "programmed", but no this behaviour wasn't programmed intentionally, it is an emergent property arising from system constraints.
Maybe the database connection pool is maxed out and the connections are saturating. Maybe some database configuration setting is too small or the server has too few file handles - whatever.
Whatever the root cause (even though that cause incidentally was implemented by a human if you trace the causal chain back far enough) this behaviour is an almost incidental unintended side effect of that.
A machine learning system is like that, but more so.
An LLM, say, is "parsing" language in some sense, but ascribing what it is doing to human design is pretty indirect.
In a way you typing words at me has in some way been "programmed" into you by every language interaction mankind has had with you.
I guess you could see it that way, but I don't think it's a particularly useful point of view.
In the same way an LLM has been in directly "programmed" via it's model architecture, training algorithm and training data, but we are nowhere near the understanding of the process to be able to consider this "programming" it yet.
This is different from a bug or hitting an unknown limitation—the selling point of this was “it makes shit up” and they went “yeah, cool, let’s have it speak for us”.
Its behavior incorporates randomness and is unpredictable and hard to keep within bounds on purpose and they decided to tell a computer to follow that unpredictable instruction set and place it in a position of speaking for the company, without a human in between. They shouldn’t have done that if they didn’t want to end up in this sort of position.
We agree that this is an engineering failure - you can't deploy an LLM like this without guardrails.
This is also a management failure in badly evaluating and managing the risks of a new technology.
We disagree in that I don't think that its behaviour being hard to predict is on purpose: we have a new technology that shows great promise as tool to work with language input and outputs. People are trying to use LLMs as general purpose language processing machines - in this case as chat agents.
I'm reacting to your comment specifically because I think you are evaluating LLMs using a mental model derived from normal software failures and LLMs or ML models in general are different enough to make that model ineffective.
I almost fully agree with your last comment, but the
reflects what I think is now an unfruitful model.
Before deploying a model like this you need safeguards in place to contain the unpredictability. Steps like the following would have been options:
* Fine-tuning the model to be more robust over their expected input domain,
* Using some RAG scheme to ground the outputs over some set of ground truths,
* Using more models to evaluate the output for deviations,
* Business processes to deal with evaluations and exceptions, Etc
Chatbots aren't people and people are actually separate legal entities responsible for their own actions.
People working for companies are sometimes separate legal entities responsible for their own actions, and sometimes they act on behalf of the company they work for and the company is responsible for their actions. It depends.
A chatbot(computer) cannot be responsible for their own actions so the only half of the coin you have left is "the company is responsible for their actions."
Computers are inanimate objects and are not recognized as legal entities.
The legal system recognizes that people, or groups of people, are subject to legal authority. This is a story about a piece of software Air Canada implemented which resulted in them posting erroneous information on their website.
Law and justice is not like a computer program that you can exploit and control without limits by being a hacker.
If the chatbot told them that they'd get a billion dollars, the courts would not hold Air Canada responsible for it, just as if a programmer put a decimal wrong and prices became obviously wrong. In this case, the chat bot gave a policy within reason and the court awarded the passenger what the bot had promised, which is a completely correct judgement.
This argument seems overly dramatic and distorted. Yes, in an outrageous situation like a billion-dollar mishap, most people would know something isn't right. But for a policy that appears legitimate, especially when it's replacing a human customer service rep, it's not that obvious. In these cases, Air Canada should definitely be held accountable.
Yes, that's exactly what I'm saying as well. Especially since they had already taken the customer's money.
Likely because the claim was considered to be within the reasonable expectations of real policy.
There is a common misconception about law that software engineers have. Code is not law. Law is not code. Just because something that looks like a function exists, you can't just plug in any inputs and expect it to have a consistent outcome.
The difference between these two cases is that even if a chat bot promised that, the judge would throw it out, because it's not reasonable. Also, the firm would have a great case against at least the CS rep for this collusion.
If your friend of a CS agent promised you a bereavement refund (As the chatbot did), even though it went against company policy, you'd have good odds of winning that case. Because the judge would find it reasonable of you to believe and expect that after speaking to a CS rep, that such a policy would actually get honored. (And the worst that would happen to the CS rep would be termination.)
Your example is significantly different.
The chatbot instructed the passenger to pay full price for a ticket but stated they could get a refund later. That refund policy was a hallucination. The victim her just walked away with a discounted ticket as promised not a billion dollars.
No, it is more similar to Air Canada hiring a monkey to push buttons to handle customer complaints. In that case, the company knows (or should know) that the given information may be wrong, but accepts the risk.
What you are describing is 1) fraud, 2) conspiracy, and 3) not a policy that a reasonable person would take at face value.
It is very different than if an employee were to, in writing, make a statement that a reasonable person would find reasonable.
Weird straw man...
So replacing all their customer support staff with AI that misleads customers is OK? That's pants on head insane, so why spend time trying to justify it.
The legal concept is called "Apparent authority". The test is whether "a reasonable third party would understand that an agent had authority to act".
("Chatbot says you can submit a form within 90 days to get a retroactive bereavement discount" sounds perfectly reasonable, so the doctrine applies.)
https://en.wikipedia.org/wiki/Apparent_authority
One major difference is the AI wasn't your friend, another is that you didn't get it hired at Air Canada, another is that the promise wasn't $1B, etc...
No, if you conspire with your friend to get them to tell you an incorrect policy, then you have no reasonable expectation that what they tell you is the real policy. If you are promised a billion dollars even without a pre-existing relationship with the agent, you have no reasonable expectation that what they are promising is the real policy because it's an unbelievably large amount.
If you are promised something reasonable by an agent of the company who you are not conspiring with, then the company is bound to follow through on the promise because you do have a reasonable expectation that what they are telling you is the real policy.
Your friend is not trained by Air Canada. The bot is Air Canada property.
If they decide it is reliable enough to be put in front of the customer, they must accept all the consequences: the benefits like having to hire less, and the cons, which is that they have to make it work correctly.
Otherwise, woopsy, we made our AI handle our accounting and it cheated, sorry IRS. That won't fly.
The claim is so outrageous that I wish there were a way (I assume there probably isn't) for the company or the lawyers to have been sanctioned outside what the plaintiff was asking for.