Air Canada
Self-service support chatbot answering fare policy questions to deflect calls from the call center
In February 2024, a Canadian tribunal ruled against Air Canada over false information given by its chatbot, establishing that a brand is responsible for what its chatbot says in the same way as for a page on its site.
Objective
Reduce call center load and cost per contact by automating answers to travelers' frequently asked questions.
The deployment
Air Canada deployed a chatbot on its public site to answer customer questions in natural language, notably about fare policies. On November 11, 2022, a customer trying to book a last-minute flight after a death asked the chatbot about bereavement fares. The chatbot stated that he could pay full fare and then request a partial refund within 90 days of the flight - an invented policy, contradicted by the official page on the same site. The customer paid, requested the refund, and Air Canada refused. After months of back-and-forth, he filed with the Civil Resolution Tribunal of British Columbia. On February 14, 2024, the tribunal ruled against Air Canada for negligent misrepresentation, explicitly rejecting the argument that the chatbot was a separate legal entity. It became the leading global legal precedent on brands' liability for what their chatbots say.
Results Proof A
Published, publicly available court decision (Moffatt v. Air Canada, 2024 BCCRT 149, CanLII): facts, timeline, amounts, and reasoning established by a tribunal, plus law firm analyses and major press.
How it works
Documented architectureThe stack in detail
- outil Chatbot propriétaire Air Canada NLP chatbot integrated into aircanada.com, generating freely on fare policies; the stack was not published.
- llm NLP conversationnel AI-assisted conversational technology; neither the model nor the vendor was made public during the litigation.
- infra Site aircanada.com (pages de politiques) The same site hosted the official fare policy, which the chatbot did not consult (no grounding), the source of the contradiction that was adjudicated.
Post-mortem
GraveyardWhat happened sourced
Nov 11, 2022: the chatbot tells J. Moffatt that a bereavement fare refund can be requested up to 90 days after the flight. Nov 2022 - Feb 2023: Air Canada refuses the refund (the actual policy excludes requests made after travel) and offers a 200 CAD credit. The customer files with the BC Civil Resolution Tribunal. Feb 14, 2024: the Moffatt v. Air Canada decision, 2024 BCCRT 149 - the tribunal finds that Air Canada 'did not take reasonable care to ensure its chatbot was accurate' and orders it to pay 650.88 CAD plus 36.14 CAD in interest and 125 CAD in fees. Press coverage notes that the chatbot was no longer accessible by the time the case drew attention.
Reason for failure sourced
The chatbot provided false information that contradicted the official policy published on the same site. The tribunal held Air Canada responsible for 'all the information on its website, whether it comes from a static page or a chatbot,' and rejected the defense that presented the chatbot as a separate entity, calling it a 'remarkable submission.'
Cost sourced
Direct: 812.02 CAD. Indirect: global reputational cost (CBC, BBC, Wired, Forbes) and the creation of a legal precedent cited internationally - the real cost is in case law.
Warning signs inferred
The contradiction between the chatbot and the official page was detectable through a simple QA test cross-checking the bot's answers against the published policies - financially sensitive topics should have been a priority test set. The absence of a disclaimer and of human escalation on sensitive fare questions was a signal. Finally, the handling of the dispute (refusal, a 200 CAD credit, silence) turned a 650 CAD incident into a global precedent.
Lessons in hindsight inferred
(1) Legally, your chatbot IS your site: everything it says binds you. (2) Grounding is mandatory: a customer bot must answer from versioned policy documents (constrained RAG plus citations), never through free generation on financially sensitive topics. (3) Plan a make-good path: honoring the bot's errors below a threshold costs less than litigation. (4) Adversarial QA on refund policies before and during production.
Yes. The customer service chatbot remains broadly valid and widely deployed (Klarna claimed the equivalent of 700 agents in 2024). The failure condemns one execution: ungrounded generation on contractual policies, with no guardrails and no remediation strategy. It sets the liability framework within which the pattern must now operate.
How your customers perceive this type of use
Sourced studiesLes consommateurs n'acceptent pas les chatbots par defaut : 64% prefereraient que les entreprises n'utilisent pas d'IA dans leur service client (Gartner, 2024) et pres d'un utilisateur sur cinq du service client par IA n'en retire aucun benefice (Qualtrics, 2025). L'acceptation se construit sur trois conditions mesurees par Salesforce : savoir qu'on parle a une IA, pouvoir escalader vers un humain, comprendre la logique de l'agent.
Acceptance conditions
- Etre informe qu'on parle a une IA et non a un humain (pres de 75% le demandent, Salesforce 2024)
- Un chemin d'escalade clair vers un agent humain (45% plus enclins a utiliser l'agent IA, Salesforce 2024)
- Une logique de l'agent clairement expliquee (44% plus enclins, Salesforce 2024)
Red lines
- Rendre l'humain injoignable : c'est la premiere inquietude des consommateurs sur l'IA dans le service client (Gartner 2024) et 50% craignent que l'IA les coupe du contact humain (Qualtrics 2025)
- Remplacer le service client par l'IA sans alternative : 53% envisageraient de partir chez un concurrent (Gartner 2024)
Sources: Salesforce 2024 · Gartner 2024 · Qualtrics 2025
The same trap, elsewhere
Sources
- S1 Moffatt v. Air Canada, 2024 BCCRT 149 (CanLII) Primary archive pending
- S2 Moffatt v. Air Canada: A Misrepresentation by an AI Chatbot - McCarthy Tétrault Established press archive pending
- S3 BC Tribunal Confirms Companies Remain Liable for Information Provided by AI Chatbot - American Bar Association Established press archive pending
- S4 Airline ordered to compensate a B.C. man because its chatbot provided inaccurate information - Dentons Data Established press archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.