The Air Canada AI chatbot misinformation case — who is responsible?
Contents
In the winter of 2022, Air Canada's website chatbot told a passenger, "If you book at full fare and apply for the bereavement fare within 90 days, you can get the difference refunded." That was wrong. Air Canada argued the chatbot was "a separate entity responsible for itself," but a Canadian tribunal didn't accept it.
At a glance
| Item | Details |
|---|---|
| Company/organization | Air Canada |
| Industry | Airlines |
| AI in dispute | Website customer-service chatbot |
| Timing of incident/ruling | Inquiry: November 2022 · Ruling: February 14, 2024 |
| Key outcome | Civil Resolution Tribunal (CRT) held the company liable for the chatbot's misinformation and ordered compensation |
Background — what went wrong
Passenger Jake Moffatt needed to book a flight urgently in November 2022 to attend his grandmother's funeral. He asked Air Canada's chatbot how to apply the bereavement fare (a discounted fare for passengers who have lost a family member). The chatbot told him he could book at full fare and get the difference refunded if he applied within 90 days, and according to CTV News, he bought the ticket trusting that answer. But another page on the same website already stated the opposite — that bereavement fares could not be claimed after travel was completed.
The core legal dispute
Moffatt requested a refund of the difference but was denied, and filed a claim with British Columbia's Civil Resolution Tribunal, which handles small claims. Air Canada argued that "the company cannot be held responsible for information provided by the chatbot." According to an analysis by McCarthy Tétrault, this amounted to arguing the chatbot should be treated as "a separate legal entity responsible for its own actions."
Tribunal member Christopher Rivers called this a "remarkable submission" and rejected it. The logic was simple: even though the chatbot has a conversational interface, it is still part of Air Canada's website, and whether information comes from a static page or a chatbot response, it is all a "representation" the company is responsible for. The tribunal found negligent misrepresentation, based on the duty of care arising from the commercial relationship between an airline and its passengers, for failing to reasonably ensure the accuracy of the chatbot's responses.
Outcome — compensation and fallout
The tribunal ordered Air Canada to pay a total of CAD 812.02, combining the fare difference, interest, and tribunal fees. Fox Business and others reported this total. The amount is small, but the case has been widely cited for establishing the principle that "if an AI says something wrong, the company that deployed it is responsible." Air Canada said it would comply with the ruling, closing the case.
The tribunal's core reasoning for finding liability
- A chatbot is a website feature, not a legal person. Being conversational doesn't change where responsibility lies.
- Companies must keep information consistent and accurate across every channel. One page being correct doesn't excuse an error on another channel.
- A disclaimer must come first to claim an exemption. McMillan LLP notes the outcome might have differed had there been advance warning that "responses may be inaccurate."
Limits of this ruling — limited weight as precedent
Analysis: This ruling came from an administrative tribunal (the CRT) that handles small claims, not a regular court. It doesn't bind other courts, and the damages were modest. McCarthy Tétrault points out that the tribunal didn't explain its finding of negligence in depth, and the chatbot's training and verification methods were never examined as an issue. Even so, experts cited by CTV News believe courts are likely to reference the same reasoning in similar future cases, even without binding force.
Applying this to your organization
- Large enterprises: Chatbot answers should be tied to a real-time policy database. A manually maintained, separate knowledge base will lag behind the chatbot whenever policy changes.
- Small and medium businesses: Questions involving money — refunds, discounts, compensation — are safer handled by a human agent than by automated chatbot responses.
- Startups: As Cloud Security Alliance notes, build in a disclaimer that responses "may not be accurate" and a system for humans to regularly review anomalous responses from the design stage onward.
References
Contents
Related posts

Call center AI adoption, why does it end in rollback — failure patterns and how to avoid them
Based on a recent survey showing 74% of companies roll back call-center AI chatbots, we contrast Klarna’s failure without escalation design and Carmoola’s success with preset criteria, distilling three recurring causes.

Why AI pilots stall before enterprise scale — five recurring traps
88% of companies use AI, but only ~6% report enterprise-wide financial impact. McKinsey, MIT, and Gartner flag five traps blocking pilots: data quality, process misfit, lagging governance, legacy integration, and change management.

Why do AI pilots stall before enterprise-wide scaling — five recurring pitfalls
88% of firms use AI, but only about 6% report enterprise-wide financial impact. McKinsey, MIT, and Gartner cite five traps: data quality, process misfit, lagging governance, legacy integration, and change management.

AI productivity paradox — work feels faster, so why don’t the numbers move?
There’s ample evidence that AI speeds up specific tasks, so why don’t company- and economy-wide productivity stats show it? From Solow’s paradox onward, we examine why this gap persists using recent CEO and research data.