Are You Liable for What Your Chatbot Says? The Air Canada Case Explained

A tribunal held Air Canada liable when its chatbot misstated a fare policy. Here is what Moffatt v. Air Canada means for small businesses, and how to stay safe.

Zalena Team6 min read

In this post
  1. What actually happened in the Air Canada case?
  2. Why does this worry small businesses more than big ones?
  3. What was the real cause?
  4. How do you make this failure mode nearly impossible?
  5. A short checklist before you turn any bot loose
  6. Which industries carry the most chatbot liability risk?
  7. Case study: Walee Fragrance, Markham (perfume ecommerce)
  8. FAQ
  9. Sources
Same technology, two outcomes. An ungrounded bot invents a policy and the business is liable. A grounded bot answers from your facts and the risk is contained.
Open the detailed diagram

A real ruling made business owners nervous about AI. Here is what actually happened, and how to make that failure mode nearly impossible.

Note

This article is general information, not legal advice. For your situation, consult a qualified lawyer.

Quick answer

Yes, a business can be held responsible for what its chatbot says. In Moffatt v. Air Canada (2024 BCCRT 149), a tribunal found Air Canada liable after its website chatbot misstated its bereavement fare policy, and rejected the airline’s argument that the chatbot was a separate legal entity. Damages were small (about CA$650), and this was a small-claims decision, not a binding precedent. Its reasoning is still worth taking seriously: “the bot said it, not us” did not work. The fix is not to avoid AI. It is to use an assistant built not to invent answers.

What actually happened in the Air Canada case?

In November 2022, Jake Moffatt asked Air Canada’s website chatbot about bereavement fares. The bot said a reduced rate could be claimed retroactively, which contradicted Air Canada’s real policy. When the refund was denied, Moffatt took it to British Columbia’s Civil Resolution Tribunal. The tribunal member found Air Canada’s claim that the chatbot was responsible for its own actions to be “a remarkable submission,” noting the chatbot “is still just a part of Air Canada’s website.” The tribunal found negligent misrepresentation, so the ruling was about a careless statement, not about enforcing a promise. The airline was ordered to pay CA$650.88 in damages (about CA$812 with interest and fees). It was a small-claims decision, and the tribunal says even its noteworthy decisions are not precedent-setting.

Why does this worry small businesses more than big ones?

A multinational can absorb a bad promise and a bruising news cycle. A small business often cannot. When you are your reputation in a local market or a niche online category, one invented policy you have to honor, or publicly walk back, does outsized damage. The caution is rational. The mistake is concluding that the answer is to avoid AI entirely.

What was the real cause?

Nobody outside Air Canada knows. The airline gave the tribunal no information about the nature of its chatbot, so the cause was never established. What is on the record is the result: a confident answer that contradicted the company’s own policy page. Whether a bot can do that depends largely on how it is built. Not every product sold as a chatbot is built the same way, and how much freedom it has to improvise is the thing to check. Chatbot vs. AI Agent vs. Conversational AI maps the generations of business AI and what each one is allowed to do on its own.

How do you make this failure mode nearly impossible?

A safer assistant is built to do the opposite of freelancing:

  • It answers only from your own information, and when it does not know, it says so instead of inventing.
  • It will not quote prices, terms, or policies it cannot verify, the kind of mistake that put Air Canada in front of a tribunal.
  • It escalates anything sensitive, unusual, or legally consequential to a human before it becomes a promise.
  • It keeps a record of what was said, so nothing happens in the dark.

That is the principle behind Zalena: grounded in the information you give it, built to hand off when a call needs judgment. The point is not that mistakes become impossible. It is that a wrong answer about your policies becomes much less likely.

A short checklist before you turn any bot loose

Is it grounded (answers only from your information)? Does it admit what it does not know? Can it quote prices or policies it cannot verify (it should not)? Does it escalate sensitive calls cleanly? Is there a record? A bot that clears these is not a liability waiting to happen. It is a tool with guardrails.

Which industries carry the most chatbot liability risk?

Anywhere a wrong answer creates a financial or legal obligation:

  • Online retail and ecommerce: returns, shipping timelines, authenticity, discount claims.
  • Travel, hospitality, and events: fares, cancellations, refund rules (the Air Canada scenario itself).
  • Insurance and financial services: coverage, eligibility, and pricing statements.
  • Healthcare, dental, and clinics: anything edging toward medical advice.
  • Real estate and property management: availability, terms, disclosures.
  • Automotive sales and service: quotes, warranties, trade in promises.

Case study: Walee Fragrance, Markham (perfume ecommerce)

Walee Fragrance Company is a Shopify based perfume and attar retailer in Markham that ships across Canada and internationally, with both a storefront and a customer line. An online store fields a constant stream of questions across time zones: “Is this alcohol free?”, “When will my order arrive?”, “What is your return policy?”, “Is this genuine?” Answered wrong, those are not just unhappy customers. A bot that invents a shipping window or a return rule can create exactly the kind of exposure the Air Canada ruling warns about.

Zalena answers Walee’s customers only from Walee’s real product data and published policies, on the phone, chat, WhatsApp, Instagram, and SMS, after hours included. She confirms what she knows (ingredients, sizes, published shipping and return terms), refuses to improvise on anything she cannot verify, and routes edge cases to a person. The result is faster answers for international shoppers at 2 a.m., and a customer facing assistant that does not make promises Walee cannot keep.

FAQ

Can a company be held liable for what its chatbot says?
Yes. In Moffatt v. Air Canada (2024), a tribunal held the company responsible for inaccurate information its chatbot provided, and rejected the idea that the chatbot was a separate entity.

How much did Air Canada have to pay?
About CA$650.88 in damages, with the total award near CA$812 including interest and fees. The financial hit was small. The reasoning is the takeaway, although this was a small-claims decision and not a binding precedent.

Should small businesses avoid AI chatbots because of this?
No. Use a chatbot that answers only from your verified information, will not invent policies, and escalates sensitive questions to a person.

What makes an AI assistant safe for customer service?
Grounded answers (no guessing), no unverifiable promises, clear human escalation, and a record of interactions.

Part of our series on small business phone strategy. Start with the overview, Press 1 for English: What My Insurance Company Taught Me About AI.

Sources

  • Moffatt v. Air Canada, 2024 BCCRT 149 (Civil Resolution Tribunal of British Columbia).
  • CBC News, coverage of the ruling.
  • American Bar Association, Business Law Today, analysis of company liability for chatbot information.

Get your nights back.

Website chat is free up to 100 messages a month.

Try it free

Or call and hear Zalena yourself: +1 (437) 524-1147