
IN this article
Explore how AI support agents enhance customer service by reducing response times and improving efficiency through automation and predictive analytics.
If your chatbot tells a customer something wrong, can you say "the bot made a mistake, that is not our policy"? A tribunal in Canada answered that in 2024, and the answer was no. For anyone running AI in customer support, that case is the clearest signal yet of where responsibility sits.
TLDR:
In Moffatt v. Air Canada (2024 BCCRT 149, February 2024), Air Canada was held responsible for its chatbot's wrong answer about bereavement fares and ordered to pay about C$812.
The tribunal rejected the idea that the chatbot was a separate entity. The company is responsible for everything on its website, including the bot.
The highest-risk answers are about money, refunds, eligibility, policy exceptions and anything regulated.
Reduce risk with grounded answers, clear escalation, a full trace of every answer and regular testing. This is not legal advice; check with counsel.
What happened in Moffatt v. Air Canada
A customer asked Air Canada's website chatbot about bereavement fares after a death in the family. The chatbot said he could book a full fare and apply for the bereavement discount afterwards. That was wrong; the airline's policy page said the discount could not be claimed after travel. When the airline refused the refund, the customer went to the British Columbia Civil Resolution Tribunal.
The tribunal found the airline liable for negligent misrepresentation and ordered it to pay the difference, about C$812 including interest and fees. It rejected the argument that the chatbot was responsible for its own actions, and said the customer had no reason to think one part of the website was more reliable than another. The amount was small. The principle was not.
What it means for your support team
The practical lesson is simple: treat what your AI agent says as what your company says. Courts and regulators are unlikely to accept "the model made it up" as a defence, and customers will not either. Where the agent states a policy, quotes a price or promises an outcome, assume you may have to honour it.
Where the risk is highest
Topic | Why it is risky |
|---|---|
Refunds, credits and fees | Wrong answers turn directly into money owed |
Eligibility and policy exceptions | The agent may promise something the policy does not allow |
Prices and promotions | Quoted prices can be held against you |
Regulated topics (health, lending, insurance) | Errors can break sector rules as well as cost money |
Account and data changes | A wrong action can harm the customer and expose data |
Controls that reduce liability
Ground every answer. The agent should answer only from your approved policies and knowledge, and say it does not know when the answer is not there.
Keep policies consistent. The Air Canada bot contradicted the airline's own policy page. Find and fix conflicting articles before the agent uses them. Fini's Knowledge Atlas flags conflicting articles for this reason.
Escalate high-risk topics. Send policy exceptions, disputes and regulated questions to a human, with the full conversation attached.
Keep a full trace. Record the sources, rules and actions behind every answer so you can see what was said and why. Fini's AI Steps trace shows the sources, the rule path and every action with inputs and results.
Disclose AI use. Tell customers they are talking to AI. It is required in a growing number of places; see our guide to AI chatbot disclosure laws.
Test before and after every change. Run fixed test sets on your riskiest topics and rerun them after every update. See regression testing AI support platforms.
Accuracy is a legal issue now
After this case, accuracy is not only a CSAT number. It is part of your risk posture. Ask vendors how they ground answers, how they handle conflicting content and what record they keep of each answer. Fini delivers 90% resolution at 99% accuracy across more than 3M resolutions a month, is SOC 2 Type II, ISO 27001:2022 and PCI DSS Level 1 certified, HIPAA-compliant and BAA-eligible, and covers GDPR and CCPA. For regulated teams, see AI agents for regulated customer support and audit-ready enterprise AI support platforms.
Final thoughts
Your chatbot speaks for your company. Keep it grounded in consistent policies, send risky topics to people, keep a record of every answer and test it often. Then talk to your counsel about how the rules apply in your markets. Fini starts at $0.49 per resolution, with no seat fees and no named plans. Every customer gets the 90-Day Money-Back Guarantee: 90% resolution in 90 days, or you get your money back (guarantee terms).
Frequently Asked Questions
Is a company liable for what its AI chatbot says?
Generally, yes. In Moffatt v. Air Canada (2024), the tribunal held the airline responsible for its chatbot's wrong answer and rejected the idea that the bot was a separate entity. Check how this applies to you with counsel.
What did Air Canada have to pay?
About C$812, covering the fare difference plus interest and fees. The amount was small, but the case set a clear example that companies answer for their chatbots.
How can we reduce legal risk from an AI support agent?
Ground answers in approved, consistent policies, escalate high-risk topics to humans, keep a full trace of every answer, disclose AI use and test before and after every change.
Is this guide legal advice?
No. It explains public cases and general practice as of November 2026. Speak to your legal team about your situation.
Related guides
Explore the guide topics to find more reading.
Co-founder

