The Assay · 11 November 2026

Who is responsible for what your chatbot says? What the Air Canada case shows

When AI answers your customers, your business has answered them. That is my reading of a 2024 Canadian tribunal decision, which held an airline responsible for what its website chatbot told a customer, after the airline argued it could not be held liable for the chatbot’s answer.

Before AI answers anyone in your name, one named person should own what it says and be able to switch it off.

Say your website has a chat box now. Or your team lets AI draft the replies to customer emails, and on a busy afternoon some of them go out unread.

It answers at eleven at night. It never keeps anyone waiting. Most of the time it is right.

Then one evening it tells a customer something your terms don’t say.

Whose answer was that?

In February 2024 a small-claims tribunal in British Columbia had to decide that question about Air Canada. The decision is public, short, and worth reading in full.

What the Air Canada chatbot decision records

In November 2022, after their grandmother died, a passenger went to Air Canada’s website to book a flight from Vancouver to Toronto at the airline’s reduced bereavement fare, and asked its chatbot how that worked.

The chatbot told them they could travel first and apply for the reduced rate afterwards, within 90 days of the ticket being issued. The tribunal had the screenshot.

The same answer linked to Air Canada’s own “Bereavement travel” page. That page said the opposite: no bereavement requests once travel was over.

The passenger flew, applied well inside the 90 days, and was told by Air Canada staff that claims after travel were not allowed. They chased the airline through December and into February. On 8 February 2023 an Air Canada representative admitted the chatbot had given “misleading words” and said the issue had been noted so the chatbot could be updated.

What the airline argued, and what the tribunal said

Air Canada argued it could not be held liable for information given by its agents, servants or representatives, including a chatbot.

The tribunal’s answer: “In effect, Air Canada suggests the chatbot is a separate legal entity that is responsible for its own actions. This is a remarkable submission.”

Then the line that matters: “It should be obvious to Air Canada that it is responsible for all the information on its website. It makes no difference whether the information comes from a static page or a chatbot.”

The claim was for negligent misrepresentation. The tribunal found that Air Canada had not taken reasonable care to make sure its chatbot was accurate, and that the passenger had been reasonable to rely on it.

It also noted that Air Canada had not explained why a customer should have to check one part of its website against another.

The airline was ordered to pay just over 800 Canadian dollars, most of it the difference between what the passenger paid and what the bereavement fare would have cost.

A small sum. The argument it lost matters more than the money.

What the decision leaves unknown

Some of the coverage went further than the decision. Several headlines called it an AI that made a policy up. The decision doesn’t say what kind of chatbot it was; Air Canada gave the tribunal no information about that. Nor does it say who inside the airline owned the chatbot, or how many other customers got the same answer.

It is not UK law. It is one small-claims decision in one Canadian province, and I’m not offering it as a rule for your business.

My reading: a chatbot nobody was standing behind

From here on it is my reading, not the tribunal’s finding.

Look at what the airline did not do. It did not describe its own chatbot to the tribunal. Its chatbot and its policy page said opposite things, and in the record the airline admits the chatbot’s “misleading words” only in reply to a customer’s complaint.

The fix on offer was to note the issue so the chatbot could be updated, after the event.

In front of the tribunal, one of its arguments was that it could not be held liable for what the chatbot said.

None of that is about the technology being bad. It reads like a tool nobody was standing behind.

An airline has layers of departments between the chat box and anyone who could say “that’s wrong, switch it off”. Between your chat box and you there may be one person, or nobody.

That cuts both ways. In most owner-run businesses the person who should own those answers is easy to find. It is just as easy for your team to assume it is someone else, or the supplier, or that you have it covered.

Put it in your own terms, with an example I have made up. Your chat box tells a customer they can return a made-to-measure worktop within 30 days. Your terms say made-to-measure items can’t be returned.

The worktop is already cut. You are now talking to a customer who has the answer in writing, and saying it was the chat box’s answer, not yours, is the argument the airline lost.

In a small business, that is what human oversight of AI comes down to: a name behind each answer.

Three questions before AI answers your customers

If AI writes to your customers in your business’s name, in a chat box, an email draft that goes out unread, or a quote, ask three things.

  1. Who, by name, owns what it says? Not the supplier. Not your IT provider. A person who would say “that answer was ours”.
  2. When did that person last put its answers to your three most common customer questions next to your own terms? Do they match?
  3. Who can switch it off, and how quickly, if it starts saying something wrong on a Saturday?

A good answer to the first is a first name, and that person knows it is theirs. “The supplier” is not an answer; the decision doesn’t mention who built the chatbot.

The second you can do this afternoon. Type your three most common customer questions into your own chat box and put its answers next to your terms.

The third is about the chat box. For email drafts the switch is simpler: someone reads them before they go.

If the answer to the first is “nobody” or “the supplier”, you have found the gap before a customer did. That is the cheap way to find it.

Of the three, the second is the one I’d most like to hear about, whatever it turned up.

Sources: Air Canada (bereavement fare chatbot), 2024 BCCRT 149, Civil Resolution Tribunal of British Columbia, small claims, decision issued 14 February 2024, https://decisions.civilresolutionbc.ca/crt/crtd/en/item/525448/index.do. Not legal advice.

Previously in The Assay: Is AI making your business more dependent on you? Look past turnover per head

The Assay is written by Rudi Prenzlin at Scalable Growth, a UK practice that gives owner-managed businesses a fractional CFO and the operational structure to run without every decision going through the owner.

← All notes Begin a conversation