In recent years, artificial intelligence has become part of customer service. Airlines, banks, shopping platforms and other companies increasingly use chatbots to answer customer questions. These systems are designed to provide information quickly without requiring a customer to wait for a human representative.
However, an important question arises when an automated system gives incorrect information: who is responsible for the mistake?
A real case involving Air Canada provided one answer to that question.
In 2022, a customer named Jake Moffatt used Air Canada's website chatbot while trying to arrange travel after the death of his grandmother. The chatbot gave him incorrect information about Air Canada's bereavement fare policy. Moffatt relied on the information and later sought compensation from the airline.
The dispute eventually went before the Civil Resolution Tribunal of British Columbia. On February 14, 2024, the tribunal released its decision in *Moffatt v. Air Canada, 2024 BCCRT 149*.
The decision became an important real-world example of the risks associated with relying on automated systems for customer information.
The Bereavement Fare Question
Air Canada had a bereavement travel policy for eligible passengers dealing with the death of an immediate family member. The policy contained specific requirements concerning the fare and the circumstances in which it could be used.
In November 2022, Jake Moffatt needed to travel following the death of his grandmother. He visited Air Canada's website and interacted with its chatbot to obtain information about bereavement fares.
According to the tribunal's decision, the chatbot provided information indicating that Moffatt could book a regular flight and then apply for a bereavement fare after the flight had been booked.
This information was incorrect.
Air Canada's actual policy did not provide the retroactive option described by the chatbot.
Moffatt relied on the chatbot's response and booked his flight.
The difference between what the chatbot told him and what Air Canada's actual policy allowed eventually became the central issue of the dispute.
The Information Was Available on Air Canada's Website
One of Air Canada's arguments was that the correct bereavement policy was available elsewhere on its website.
The company argued that Moffatt could have found the correct information by checking the relevant webpage rather than relying on the chatbot.
The tribunal did not accept this argument as a reason to avoid responsibility.
The tribunal specifically considered whether customers should be expected to verify information given by one part of a company's website against another part of the same website.
The tribunal found that Air Canada had not explained why its bereavement webpage should automatically be considered more trustworthy than its chatbot.
The tribunal also stated that customers should not have to double-check information from one part of the company's website against another part simply because the first source was a chatbot.
This point was central to the decision.
The chatbot was not an independent website operated by someone unrelated to Air Canada. It was part of Air Canada's own website.
Air Canada's Argument About the Chatbot
The case became particularly significant because of Air Canada's position regarding responsibility for the chatbot.
Air Canada argued that it should not be held responsible for information provided by the chatbot in the same way that it would be responsible for information provided by an employee or representative.
The tribunal rejected that position.
The tribunal noted that the chatbot was part of Air Canada's website. It did not treat the chatbot as a separate legal entity that could independently bear responsibility for the information it provided.
The tribunal described Air Canada's position as remarkable and concluded that the company remained responsible for the information appearing on its website.
The tribunal's reasoning was straightforward: the method through which information is delivered does not automatically remove the company's responsibility for that information.
Whether information comes from a static webpage or an interactive chatbot, the customer is still interacting with the company's official website.
The Tribunal Found a Lack of Reasonable Care
The tribunal ultimately found that Air Canada had not taken reasonable care to ensure that its chatbot provided accurate information.
This finding is significant because the case was not simply about whether a chatbot made a technical mistake.
The tribunal considered the relationship between the company, its technology and its customer.
Air Canada had chosen to place the chatbot on its website as a customer-service tool. A customer used that tool to obtain information about an airline policy. The information supplied by the chatbot was incorrect, and the customer relied on it.
The tribunal therefore held Air Canada responsible for the consequences of the misinformation.
The Financial Consequences
The case resulted in an award in favour of Moffatt.
The tribunal ordered Air Canada to pay compensation for the loss associated with the incorrect information, along with applicable interest and tribunal-related costs.
The amount involved was relatively small compared with the scale of Air Canada's business.
However, the importance of the case was not primarily the amount of money involved.
The case demonstrated that inaccurate information supplied by an automated customer-service system can have real consequences and that a company may remain responsible for the information delivered through its own technology.
Why the Case Matters
The Air Canada case became widely discussed because it involved an issue that is becoming increasingly common.
Companies are using artificial intelligence to communicate directly with customers.
A customer may not know whether a response has been written by an employee, generated automatically, retrieved from a database or produced by a generative AI system.
From the customer's perspective, the important fact is that the information came through the company's official service.
This creates a potential problem.
If a company presents a chatbot as part of its customer-service system, customers may reasonably rely on the information it provides.
The Air Canada decision demonstrated that companies cannot simply assume that calling a system a chatbot eliminates responsibility for what that system tells customers.
The Technology Behind the Problem
The incident also demonstrates one of the limitations of artificial intelligence systems.
AI systems can produce responses that appear clear and convincing while still being incorrect.
This problem is particularly relevant to generative AI. The National Institute of Standards and Technology identifies “confabulation” as a risk associated with generative AI. In this context, confabulation refers to an AI system producing false or erroneous information while presenting it as though it were correct.
This creates a special challenge for users.
A traditional computer error may be obvious because the system stops working or displays an error message.
An AI error can be different.
The system may continue functioning normally and provide a fluent, complete response. The problem may only become visible when someone checks the information against a reliable source.
The Air Canada incident demonstrates how this technical characteristic can affect ordinary consumers.
Human Trust Was an Important Part of the Incident
The customer did not interact with the chatbot as part of an experiment.
The chatbot was available through Air Canada's official website.
The purpose of using it was to obtain information about an actual airline policy.
That matters because people generally expect official company channels to provide reliable information.
The incident therefore involved more than an incorrect computer-generated answer. It involved the relationship between technology and customer trust.
When a company provides an automated system for answering customer questions, customers may reasonably assume that the company has taken steps to make sure the information is accurate.
If that information is wrong, the customer may make a decision based on it.
The Importance of Verification
The case also demonstrates why information provided by automated systems should be verified when it concerns important decisions.
For businesses, this means that AI systems should be tested before being deployed and monitored after deployment.
For customers, it means that information involving money, legal rights, travel requirements, medical matters or other important decisions should be checked against an authoritative source when possible.
However, the Air Canada decision also shows that companies cannot simply place the entire burden of verification on customers.
If a company operates multiple information channels, those channels should provide consistent and accurate information.
A customer should not have to determine which part of an official website is trustworthy.
What Businesses Can Learn
The case provides several practical lessons for businesses using AI.
First, companies should know exactly what information their chatbots are providing.
Second, important policy information should be connected to reliable and current sources.
Third, automated systems should be tested using realistic customer questions rather than only simple technical tests.
Fourth, companies should monitor chatbot responses after deployment because problems may not become visible during initial testing.
Fifth, customers should have an accessible method of contacting a human representative when a chatbot cannot provide a reliable answer.
These measures do not mean that businesses should stop using AI.
Instead, they show that AI systems need appropriate controls.
What Customers Can Learn
Customers can also learn from the incident.
A chatbot response should not automatically be treated as a guarantee, particularly when the information concerns money or important decisions.
If an answer seems unusual or conflicts with another official source, it is reasonable to seek confirmation.
At the same time, customers should be able to expect reasonable accuracy from an official company service.
The responsibility for reliable customer communication cannot rest entirely on the person using the technology.
A Broader Issue for Artificial Intelligence
The Air Canada case was about an airline chatbot, but the underlying issue is broader.
AI systems are now being introduced into many areas of daily life.
They are used for customer support, education, banking, recruitment, information retrieval and many other activities.
As AI becomes more common, people will increasingly encounter situations where they need to decide whether an AI-generated answer can be trusted.
The Air Canada case provides a practical example of why accuracy and accountability need to develop alongside AI technology.
Artificial intelligence can make services faster and more accessible, but an automated system does not automatically make information reliable.
The Air Canada chatbot case began with a customer seeking information about a bereavement fare after the death of his grandmother.
The customer used Air Canada's official website and received incorrect information from its chatbot.
He relied on that information, and the disagreement that followed eventually reached the British Columbia Civil Resolution Tribunal.
The tribunal found that Air Canada had not taken reasonable care to ensure that its chatbot was accurate. It also rejected the idea that the chatbot should be treated as something separate from the company simply because it was an automated system.
The case is important because it provides a real example of the consequences of AI-generated misinformation.
The technology itself was not the only issue. The central issue was responsibility.
When a company uses an automated system to communicate with customers, that system becomes part of the company's service. If the system provides incorrect information, the company cannot necessarily avoid responsibility by pointing to the technology.
The Air Canada case therefore offers a clear lesson for the growing use of artificial intelligence:
Automation can change how information is delivered, but it does not automatically change who is responsible for that information.
As more organisations adopt AI chatbots, accuracy, verification, monitoring and accountability will become increasingly important. The experience of Air Canada and Jake Moffatt shows why those safeguards matter in the real world.
References:
- Moffatt v. Air Canada, 2024 BCCRT 149, British Columbia Civil Resolution Tribunal, February 14, 2024.
- CanLII Blog, discussion of Moffatt v. Air Canada, including the tribunal's findings concerning Air Canada's chatbot.
- National Institute of Standards and Technology (NIST), Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile, NIST AI 600-1, 2024.