Air Canada was held liable for its chatbot's wrong answer. Here's what the ruling means, what's happened since, and a 10-point checklist for your site.
A grieving customer asked Air Canada's website chatbot about bereavement fares, got the wrong answer, and was refused the discount. When he took it to a tribunal, Air Canada argued the chatbot was, in effect, responsible for its own words. The tribunal called that "a remarkable submission" and made the airline pay.
If your company has a chat window on its website, answering questions about your products, your pricing or your policies, this is the case to know about. Chatbot liability sounds like a problem for airlines and banks. The principle the tribunal applied is much simpler than that, and it applies to every business: what your website tells a customer, you told the customer.
I build AI chat agents for a living, so I have a stake in this. Here's my honest read, with one caveat up front: I'm not a lawyer, and this isn't legal advice.
What actually happened
The case is Moffatt v. Air Canada, 2024 BCCRT 149, decided by British Columbia's Civil Resolution Tribunal on February 14, 2024, according to Law360 Canada.
After his grandmother died in November 2022, Jake Moffatt asked the chatbot on Air Canada's site about bereavement fares. The bot told him that if he had already travelled, he could submit his ticket for a reduced bereavement rate within 90 days by completing a refund form, as quoted in the decision summaries.
That wasn't the policy. Air Canada's own bereavement page said the policy didn't apply to requests made after travel was completed. And the chatbot's answer had linked to that very page.
Moffatt booked flights to Toronto and back for $1,640.36 in total, then applied for the discount. Air Canada refused, offered a $200 coupon as a goodwill gesture, and he turned it down.
Air Canada's defence had two parts. The correct policy was on its website, so he should have found it. And the chatbot was, in the tribunal's summary of the argument, "a separate legal entity that is responsible for its own actions."
Tribunal member Christopher Rivers rejected both. The chatbot, he wrote, is "still just a part of Air Canada's website," and "it should be obvious to Air Canada that it is responsible for all the information on its website." He also found there was no reason a customer should be expected to know that one part of the site was accurate and another wasn't.
He found negligent misrepresentation and ordered Air Canada to pay $650.88 in damages, $36.14 in interest and $125 in tribunal fees: $812.02 in all. Air Canada told Daily Hive it would comply and considered the matter closed.
Some fair context
It's worth keeping this in proportion. The Civil Resolution Tribunal is BC's online small-claims body. As MLT Aikins notes, its decisions carry weight but aren't binding on higher courts. The amount was small, and the decision doesn't turn on what technology was behind the bot.
But that's exactly why it travels so well. The tribunal didn't need any new AI law. It used an old idea: a business is responsible for what it tells customers on its own website. That idea doesn't care whether the words came from a static page, a staff member or a chatbot.
What's happened since the Air Canada ruling
The Moffatt decision landed in February 2024. In the months since, the pattern has repeated in public, and regulators have started writing the same idea into law. A few developments worth knowing:
- March 2024: New York City's MyCity chatbot. The city's AI assistant for business owners told users they could take workers' tips and refuse tenants with housing vouchers, both wrong, The Markup found. The site carried a disclaimer that the bot "may occasionally produce incorrect, harmful or biased content." Why it matters: a disclaimer didn't stop the story. People still read the answer as the organisation's answer.
- September 2024: the FTC's "Operation AI Comply". The US Federal Trade Commission announced a sweep against deceptive AI claims, including a $193,000 proposed settlement with DoNotPay over its "robot lawyer" marketing. Chair Lina Khan put it plainly: "there is no AI exemption from the laws on the books." Why it matters: what you claim your AI can do is held to the same standard as any other claim.
- December 2024: the EU's new Product Liability Directive. Directive (EU) 2024/2853 explicitly defines "product" to include software. It entered into force in December 2024, member states must transpose it by 9 December 2026, and it applies to products placed on the market after that date. Why it matters: software suppliers, including AI vendors, face clearer liability rules for defective products in the EU.
- April 2025: Cursor's support bot invents a policy. A user asked why they kept being logged out when switching devices. An AI support agent called "Sam" replied that it was "expected behavior" under a new login policy. No such policy existed, users cancelled subscriptions, and a co-founder apologised for an "incorrect response from a front-line AI support bot." Why it matters: it's the Air Canada story again, this time at a software company, with the cost paid in lost customers rather than a tribunal award.
- August 2026: the EU AI Act's transparency rules. The European Commission says that for chatbots, "humans should be made aware that they are interacting with a machine," and that the transparency rules came into effect in August 2026. Why it matters: if you serve people in the EU, telling visitors they're talking to an AI is now a legal duty, not just good manners.
Even before the ruling, the warning signs were there. In January 2024, DPD's chatbot swore at a customer and called DPD the "worst delivery firm in the world." The company blamed an error after a system update and disabled the AI element. None of these incidents involved a big court case. All of them landed on the company that ran the bot.
Why chatbot liability matters for your business
Picture a prospect on your site at 10pm. "Do annual contracts come with a 90-day free trial?" Your chatbot, trying to be helpful, says yes. You've never offered that. The prospect screenshots it, signs up, and on Monday someone on your sales team has to explain that the website was wrong.
Now you're choosing between honouring a promise you never made and arguing with a new customer about your own chat window. The Air Canada case suggests "the bot said it, not us" isn't a strong position. And look at the detail that hurts most: Air Canada's bot linked to the correct page. Having the right policy somewhere on your site didn't save it.
No company can have its legal team check every answer the chat gives. So the practical question is: how do you keep an AI's words and actions inside what you actually offer?
A 10-point checklist for your website's chatbot
Whatever tool you use, ours or anyone else's, these are the habits I'd put in place before a chatbot speaks for your business:
- Put your real policies in its knowledge base. Pricing, refunds, trials, warranties, cancellation terms. If the bot can't find the policy, it will fill the gap.
- Keep one source of truth. When a policy changes on your website, update the chatbot's content the same day. Air Canada's bot and its policy page disagreed.
- Test the risky questions before launch. Write down 20 questions where a wrong answer costs money: discounts, refunds, deadlines, guarantees, eligibility. Ask them, and ask them in different ways.
- Tell it what not to answer. Legal advice, contract changes, custom pricing and exceptions should go to a person, not get a best guess.
- Require a human for anything that commits you. Bookings, quotes, refunds and special terms should need a click from your team before they're final.
- Say it's an AI. Make it clear in the chat window. In the EU it's now required, and everywhere else it sets fair expectations.
- Make it easy to reach a person. A visible route to a human, by chat handover, email or phone, stops a frustrated customer from relying on a guess.
- Log every conversation. If a customer says "your chat told me X," you want the transcript, not a debate.
- Review wrong answers every week. Pick an owner. Read a sample of chats, fix the source content, and record what changed.
- Don't treat a disclaimer as a shield. Keep one, but plan as if every answer will be read as your company's word. That's how the tribunal read it.
How we handle it: your team stays in control
I'll be straight about the limits first. No AI chat, including ours, can guarantee every answer is correct. What you can do is shrink the room for a wrong promise, and make sure a person signs off on anything that commits your business. That's how we've built AtomicX:
- Answers come from your content. You connect a knowledge base (your website, PDFs, policy documents), and the agent answers from it rather than from the open internet. Keep your policies there, and keep them current.
- The visitor clicks Start. Before the AI runs a workflow, such as a quote request or a demo request form, the visitor has to press Start. A "yes" typed in chat isn't enough.
- Your staff Accept bookings. Every appointment made in chat starts as a request. Nothing is confirmed until someone on your team accepts it.
- You can see why it said that. Show AI Details shows the knowledge base content the AI used, its reasoning and a confidence score, so a wrong answer can be traced back to its source.
- You can Take over. Anyone on your team can step into a live chat, then Release to AI when they're done. In the mobile app, they can tick Teach the agent this answer so it doesn't repeat the mistake.
None of this makes the AI a lawyer. It means the things that commit your business, like a booking, a form submission or a policy question you're unsure about, have a human click in the path.
The bigger pattern
Air Canada's instinct was to treat the chatbot as something outside the company. I understand the temptation. These tools can feel like a third party you plugged in.
The tribunal's answer is the one I'd give my own clients: it's your website, so it's your voice. That doesn't turn every chatbot slip into a court case, but customers can reasonably expect your chat window and your policy page to agree.
The safest AI on your site isn't the cleverest one. It's the one that answers from your own content, asks before it acts, and lets your team see and fix what it said.
Air Canada chatbot ruling: quick answers
Who won the Air Canada chatbot case?
Jake Moffatt, the customer. The tribunal found Air Canada liable for negligent misrepresentation by its website chatbot and rejected the argument that the bot was responsible for its own actions.
How much did Air Canada have to pay?
$812.02 in total: $650.88 in damages, $36.14 in interest and $125 in tribunal fees.
Is the ruling binding on other courts?
No. It came from BC's Civil Resolution Tribunal, a small-claims body whose decisions aren't binding on higher courts. Its reasoning is still the clearest public statement of a simple principle: a business is responsible for what its website tells customers.
Does a disclaimer protect my business?
I wouldn't count on it. NYC's MyCity bot had one and still made headlines. Disclose that it's an AI, but rely on good content, testing and human sign-off for anything that commits you.
Want to see how your team stays in control of what the AI says and does? Book a 20-minute demo and we'll walk through it on your own site.