Air Canada loses court case over chatbot hallucination

Air Canada tried to argue in court that its customer service chatbot is an independent legal entity that should be held responsible for its own mistakes. The judge was not amused.

The dispute dates back to 2022, when passenger Jake Moffatt needed to book a last-minute flight from Vancouver to Toronto to attend his grandmother's funeral. Unsure about how bereavement travel fares worked, Moffatt asked the airline's website chatbot for guidance.

The bot gave him clear instructions: book the full-price ticket now, and apply for a refund for the price difference within 90 days. Moffatt followed the advice. But when he submitted the paperwork after the trip, Air Canada rejected his claim.

It turned out the airline's actual policy stated that bereavement rates could never be claimed retroactively once travel was completed.

A chatbot is not a person

Moffatt showed the airline a screenshot of the chat. Air Canada acknowledged the bot had used misleading words, promised to update the software, and offered a $200 travel voucher. Moffatt declined and took the carrier to British Columbia's Civil Resolution Tribunal.

In court, Air Canada mounted a defense that tribunal member Christopher Rivers described as remarkable. The airline argued the chatbot was a separate legal entity responsible for its own actions. Air Canada also claimed Moffatt should have verified the bot's words against the static bereavement policy page linked elsewhere on its website.

The tribunal rejected that logic entirely. Rivers ruled that a chatbot is simply an interactive component of the airline's website. There is no reason why a consumer should assume that static text on a corporate site is inherently more trustworthy than an automated tool living on that exact same site.

The tribunal ordered Air Canada to pay Moffatt 650.88 Canadian dollars for the fare difference, plus 36.14 dollars in interest and 125 dollars in tribunal fees.

Why it matters

This is one of the first major legal rulings to tackle corporate liability for hallucinating chatbots, and it sets a vital precedent for the age of automated customer service.

For consumers, it draws a bright line: if an enterprise puts a bot on its front door, the company owns every word that bot utters. You do not need to double-check whether the AI is lying against fine print buried three menus deep. Legally, the bot speaks for the brand.

For businesses rushing to replace human support desks with generative AI agents, this is a loud wake-up call. Air Canada tried to have it both ways, pocketing the cost savings of automation while attempting to disown the legal risk when the model hallucinated. We think that defense looked absurd from the start, and courts clearly agree.

If you deploy an agent that can make promises, calculate refunds, or interpret contract terms, you now have to assume full liability for its hallucinations. Expect legal and compliance teams to demand tighter retrieval boundaries and much faster human escalation paths before letting these bots touch real money.