What was the Air Canada chatbot ruling?
In the Air Canada chatbot ruling, a British Columbia tribunal held the airline responsible for wrong fare advice its website chatbot gave a grieving customer and ordered it to pay $812.02 in Canadian dollars. In May 2026, a German court ordered a cosmetic treatment company to stop its artificial intelligence (AI) chatbot’s false claims.
Jake Moffatt’s grandmother died in Ontario in November, four years ago. That same day, Moffatt went to Air Canada’s website to book a flight from Vancouver to Toronto. They asked the site’s chatbot about bereavement fares, the reduced fares an airline may offer after a death in the family.
The chatbot said they could fly now and apply for the reduced rate within 90 days. That is about three months. Moffatt kept a screenshot of that answer. The airline’s own policy page said the opposite. Here is what a tribunal decided, and what it means for anyone who asks a company’s chatbot a question.
Contents
What happened with the Air Canada chatbot?#
A grieving customer got a wrong answer about a discount, and the airline refused to honor it. The chatbot told Moffatt to submit the ticket for a reduced bereavement rate within 90 days of the date it was issued. That was wrong.
No person typed that answer. Both sides implicitly agreed Moffatt was not chatting with anyone who worked for Air Canada. The words “bereavement fares” linked to a separate page. That page said the policy did not apply after travel. The two answers clashed.
Moffatt trusted the chatbot. They booked a one-way flight to Toronto for $794.98, all amounts in Canadian dollars. Five days later, they booked the flight home for $845.38.
They applied for the reduced fare the next day, well inside the 90 days. Then they waited. Emails went back and forth for months. In one, Moffatt sent the screenshot and a death certificate.
In February of the following year, an Air Canada representative admitted the chatbot had given “misleading words.” The airline still refused to pay. So Moffatt took the dispute to British Columbia’s Civil Resolution Tribunal, an online tribunal that hears small claims.
Did Air Canada have to honor its chatbot’s answer?#
Yes, in effect. In February of the year before last, the tribunal ordered Air Canada to pay Moffatt $812.02. That covered $650.88 in damages, $36.14 in interest and $125 in tribunal fees. The airline had to pay what the wrong answer cost.
The math was simple. The damages were the gap between two fares. The tribunal found the bereavement fare was worth $380 each way.
Moffatt paid $1,630.36 for the two flights, in Canadian dollars. The tribunal found they should have paid $979.48. The difference is the $650.88.
The airline lost other arguments too. It said it had already given Moffatt a $200 coupon, so that amount should come off the bill. The tribunal said no. It also pointed to the terms in its tariff, the airline’s filed rules for fares. But it never handed the tribunal the part of the tariff it relied on, so that defense failed.
Can a company be held liable for its chatbot?#
Yes, in the cases decided so far. In effect, the tribunal said, Air Canada suggested the chatbot was responsible for its own actions. Tribunal member Christopher C. Rivers called that “a remarkable submission.” He wrote that the chatbot “is still just a part of Air Canada’s website.”
He added that it makes no difference whether information comes from a static page or a chatbot. The company owns the answer, whatever typed it. A chatbot on a company’s website is like a clerk at the counter. If the clerk quotes you the wrong price, the store answers for it.
The tribunal also rejected the idea that the customer should have caught the error. It asked why one page of the site should be trusted more than another. Nobody should need a second source to check the first.
A German court went further in May 2026. The Higher Regional Court in Hamm heard the case first, brought by a consumer group. A cosmetic treatment company had an artificial intelligence (AI) chatbot on its website for booking appointments and answering questions.
The chatbot said the two doctors behind the company were specialists in plastic and aesthetic surgery. They held no such titles. The claim was false. The court ordered the company to stop. The chatbot was not a “third party” the company could blame, it ruled. The company was responsible even if it had fed the chatbot only correct data.
Are chatbots legally binding?#
Not in any general sense, and both rulings are narrower than the headlines. The Air Canada case was a small claim decided by one tribunal member. The tribunal’s own decisions say earlier rulings are “not binding” on later ones, only persuasive.
The tribunal never ruled a chatbot’s answer is a contract. That matters. It found the airline was careless about what its own website told a customer. The money covered what the wrong answer cost, nothing more.
The German ruling is not final either. It can still change. The court allowed an appeal to Germany’s Federal Court of Justice, because the case raises new questions about who answers for a chatbot’s false statements.
Europe is also changing its rules on faulty products. The European Union (EU) product liability directive counts software as a product. It applies to products placed on the market after 9 December 2026. It covers death, personal injury, damaged property and destroyed data. Pure money loss does not by itself trigger it. Moffatt’s loss was that kind.
This article did not review the law in the United States. For the rules on telling people they are talking to a machine, see this site’s piece on AI disclosure requirements.
Who is liable when a chatbot gives wrong information?#
In the two rulings here, the company that ran the chatbot was liable. Neither court let the business hand the blame to the software. That does not mean every wrong answer turns into money. A claim still has to show the answer was wrong, that relying on it was reasonable, and that it cost something.
That is why the record matters. Proof wins. Moffatt’s screenshot of the chatbot’s answer was the evidence at the center of the case. A few habits help anyone who asks a company’s chatbot a question that touches money:
- Screenshot the full answer, with the date showing.
- Ask the company to confirm anything important by email.
- Save the emails and the receipts.
- Note any page the answer links to.
A screenshot works the way a receipt does. It turns a conversation into proof. The same care applies to any policy claim a company or an article puts in front of you.
Businesses are learning the same lesson from the other side. A business can check its chatbot against its own policy pages before customers do. It can fix a wrong answer the same day. Air Canada’s representative said the airline had noted the issue so it could update the chatbot.
Neither court accepted “the software said it” as a defense. A company that puts a chatbot in front of its customers owns what it says, the same as the words on its printed page. Rules that try to keep AI safe are still taking shape, as this site found in two ways to keep AI safe.
After a death in the family, Moffatt flew on one answer. The answer was wrong, and it took more than a year and a tribunal to put it right. Keep the screenshot. For a family in that moment, it was worth more than the chatbot.
Source: British Columbia Civil Resolution Tribunal, Moffatt v. Air Canada decision.
