When the chatbot is wrong, who pays?

In February 2024, a small-claims tribunal in British Columbia issued a decision that has been read far beyond the world of airline refunds. In Moffatt v. Air Canada, the British Columbia Civil Resolution Tribunal ruled that an airline was liable for inaccurate advice given by the chatbot on its own website — and rejected, in plain terms, the argument that the bot was somehow responsible for itself. For anyone deploying an AI agent that talks to customers, it is a short and unusually clear precedent.

What happened

The facts are sad and ordinary. Jake Moffatt's grandmother died on 11 November 2022. Needing to fly from Vancouver to Toronto for the funeral, he went to Air Canada's website and asked its support chatbot about bereavement fares — the discounted tickets many airlines offer to people travelling after a death in the family.

The chatbot told him he could buy a ticket at the normal price and then apply for the bereavement rate afterwards. In its own words:

If you need to travel immediately or have already travelled and would like to submit your ticket for a reduced bereavement rate, kindly do so within 90 days of the date your ticket was issued by completing our Ticket Refund Application form.

Acting on that, Moffatt booked full-fare flights — roughly CA$1,630 round trip — and, within the 90-day window the bot had described, applied for a partial refund and submitted his grandmother's death certificate. Air Canada refused. Its actual policy, set out on a separate page the chatbot had even linked to, does not allow bereavement fares to be claimed retroactively after travel. The bot had confidently described a policy that did not exist.

The dispute

When Moffatt took the matter to the tribunal, it was Air Canada's defence that made the case notable. The airline argued, among other things, that the chatbot was "a separate legal entity that is responsible for its own actions" — as if the software, and not the company that deployed it, should answer for the bad advice. It also argued that the correct information was available elsewhere on its website, so Moffatt should have cross-checked.

Tribunal member Christopher Rivers was unpersuaded. Air Canada, he found, owed Moffatt a duty of care, and the standard of care required it to take reasonable steps to ensure its representations were accurate and not misleading. The chatbot was part of Air Canada's website, and the company was responsible for everything on it — whether the words came from a static page or a chatbot. It made no sense, he reasoned, to expect a customer to know that one part of the site was trustworthy while another was not, or to go hunting for a contradictory page to double-check the answer they had just been given.

The ruling

The tribunal found Air Canada had negligently misrepresented its bereavement policy and ordered it to compensate Moffatt for the gap between what he paid and the fare he had been led to expect: CA$650.88 in damages, plus pre-judgment interest and tribunal fees, for a total of about CA$812.

The dollar figure is trivial. The principle is not. A public body looked at the now-common corporate instinct to treat an AI assistant as a quasi-independent actor — "the bot said it, not us" — and rejected it outright.

Why it matters

As companies rush to put AI agents in front of customers, Moffatt is the case their lawyers cite first. Its lesson is simple: an agent's output is your output. If a system can state a policy, quote a price, or promise a refund, the organisation behind it is on the hook for what it says — no matter how the answer was generated.

That reframes the engineering problem. The real risk in a customer-facing agent is not that it is rude or slow; it is that it will state something false with total confidence, exactly as Air Canada's did. Reducing that risk means grounding agents in the company's real, current data instead of letting them improvise; scoping what they are allowed to assert; and logging every answer, so that when a dispute arises there is a record of precisely what was said and on what basis.

It is the same principle we build Flowtly's business agents around: an agent should answer only from approved data, stay inside the guardrails of its role, and leave an audit trail behind every decision. The Air Canada ruling is a reminder of what the alternative can cost — and that "the chatbot said it" is not a defence.