Moffatt v Air Canada: the airline bound by its chatbot's invented policy

What happened

On 11 November 2022, the day his grandmother died, Jake Moffatt consulted the support chatbot on Air Canada's website about bereavement fares. The chatbot's answer, preserved in a screenshot before the tribunal, said in part: "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."

That was wrong. The words "bereavement fares" in the same answer were hyperlinked to Air Canada's actual policy page, which said the policy does not apply to requests made after travel is completed. Relying on the chatbot, Moffatt bought full-fare tickets, Vancouver–Toronto and return, totalling $1,630.36 CAD, and applied for the reduction on 17 November, well inside the 90-day window the chatbot had described.

Air Canada refused the refund. On 8 February 2023 an Air Canada representative admitted in writing that the chatbot had used "misleading words," pointed to the hyperlink, and said the issue had been noted so the chatbot could be updated. The airline offered a $200 coupon. Moffatt declined it and filed a small claim with the British Columbia Civil Resolution Tribunal.

On 14 February 2024, Tribunal Member Christopher Rivers held Air Canada liable for negligent misrepresentation and ordered it to pay $812.02 CAD ($650.88 damages, $36.14 interest, $125 fees). Air Canada told Ars Technica it would comply and considered the matter closed. Two days after the decision, the chatbot appeared to have been removed from the airline's website; Air Canada never confirmed why.

One fact the record does not establish: the chatbot's technology. The tribunal noted Air Canada "did not provide any information about the nature of its chatbot." Whether it was generative or scripted is unknown; "AI chatbot" is the media's label, not the tribunal's finding. That gap matters less than it seems, because the ruling's logic is technology-agnostic.

The duty engaged

The cause of action was the tort of negligent misrepresentation, applying the elements from the Supreme Court of Canada's decision in Queen v Cognos: a duty of care, an untrue or misleading representation, negligence in making it, reasonable reliance, and resulting loss. The tribunal found all five satisfied.

The duty arose from the ordinary commercial relationship between a service provider and a consumer: a company must "take reasonable care to ensure their representations are accurate and not misleading." The representation was the chatbot's answer. The negligence was deploying an automated channel that contradicted the company's own system of record without reconciliation or review. The reliance was reasonable. The tribunal expressly rejected the suggestion that Moffatt should have cross-checked the chatbot against the policy page it linked to: "There is no reason why Mr. Moffatt should know that one section of Air Canada's webpage is accurate, and another is not."

For UK and EU readers, the read-across is direct. The same facts in England would run through negligent misstatement under Hedley Byrne principles and the Consumer Protection from Unfair Trading Regulations 2008 (misleading actions); in the EU, the Unfair Commercial Practices Directive does equivalent work. No data-protection duty was engaged (this is a misrepresentation case, not a data incident), which is precisely why it generalises: it attaches liability to what an automated agent says, independent of any breach.

The liability chain

The tribunal's allocation is the cleanest on the public record, because Air Canada ran the argument every deployer is tempted by and lost.

The deployer (Air Canada) argued, in the tribunal's words, that it "cannot be held liable for information provided by one of its agents, servants, or representatives, including a chatbot." Member Rivers' characterisation has become the most-quoted sentence in agent liability: "In effect, Air Canada suggests the chatbot is a separate legal entity that is responsible for its own actions. This is a remarkable submission." The holding: a company is responsible for all the information on its website, whether it comes from a static page or a chatbot. Precision matters here: "separate legal entity" was the tribunal's characterisation of the submission, not Air Canada's verbatim pleading.

The vendor layer never entered the case. Whether the chatbot was built in-house or supplied by a third party is not in the public record; no vendor was named, no indemnity surfaced. The consumer-facing liability landed entirely on the deployer; any recourse against a supplier would be a separate contractual matter invisible to the customer. That is the default shape: the deployer fronts the liability, then looks to its contracts.

Two procedural details with general lessons. Air Canada relied on its Domestic Tariff as a contractual defence but never filed the tariff text; the tribunal held an asserted-but-unproduced contract proves nothing. And it produced no evidence of the actual bereavement fare, so the tribunal drew an adverse inference and accepted the claimant's figure. Companies defending agent-output claims with paperwork they decline to produce should expect the same.

A weight caveat this register will apply consistently: the CRT is an online small-claims tribunal, not a court, deciding on written submissions. The decision binds nobody. It has nonetheless become the standard citation for agent-output liability (analysed by McCarthy Tétrault, the American Bar Association, and a 2025 Singapore Academy of Law case comment) because no court has yet had to say anything better.

What would have prevented it

• Reconciliation against the system of record. The chatbot contradicted a policy page it was itself linking to. Automated answers that touch policy, price, or entitlement need to be generated from, or checked against, the authoritative source.
• Notice-and-correction discipline. Air Canada had written notice of the defect in February 2023, a year before the decision, and the public record shows no correction until the chatbot disappeared entirely. The cheapest exit was taken last.
• Scope restriction for the channel. A support bot that declines to answer fare-entitlement questions and routes to the policy page gives up some deflection rate and removes this entire liability class.
• Honest expectation-setting. Experts quoted post-decision noted a disclaimer might have changed the reliance analysis. A disclaimer is the weakest of these controls, but its total absence made reliance straightforwardly reasonable.

Mapped controls

• OWASP Top 10 for Agentic Applications 2026: no category squarely covers non-adversarial hallucination to a consumer; the nearest is ASI09 Human-Agent Trust Exploitation, whose text warns that over-reliance on confident agent output "increases the chance of harmful decision-making" and cross-references LLM09:2025 Misinformation. Mapped here as a partial analogue, flagged as such.
• Singapore IMDA Model AI Governance Framework for Agentic AI: the framework's end-user transparency and human-accountability dimensions are the controls this incident proves out — a named human owner for the channel's accuracy would have caught a policy contradiction the company had written notice of.
• NIST AI RMF: primarily a MEASURE and MANAGE failure. The chatbot's output was never validated for accuracy against the authoritative policy source (MEASURE), and a defect the company had notice of went unmanaged until the channel was removed (MANAGE).
• The general rule the case stands for: the agent's words are the principal's words. Every entry in this register involving a customer-facing agent inherits Moffatt as its baseline.

Sources

• Moffatt v Air Canada, 2024 BCCRT 149, full decision text (BC Civil Resolution Tribunal decisions database) — checked June 2026 [primary]
• CBC News, "Air Canada found liable for chatbot's bad advice on plane tickets" (Jason Proctor), 15 February 2024 — checked June 2026
• Ars Technica, "Air Canada must honor refund policy invented by airline's chatbot" (Ashley Belanger), 16 February 2024 — checked June 2026
• McCarthy Tétrault TechLex, "Moffatt v. Air Canada: A Misrepresentation by an AI Chatbot" (Barry Sookman), 19 February 2024 — checked June 2026
• American Bar Association, Business Law Today, "BC Tribunal Confirms Companies Remain Liable for Information Provided by AI Chatbot" (Lifshitz & Hung), 29 February 2024 — checked June 2026
• Singapore Academy of Law Practitioner, "Chatbots and Liability for Negligent Misrepresentation" [2025] SAL Prac 16 — checked June 2026
• OWASP Top 10 for Agentic Applications 2026 (published 9 December 2025) — checked June 2026 [primary]
• IMDA, Model AI Governance Framework for Agentic AI (January 2026) — checked June 2026 [primary]

Cite this entry as AIR-2026-003 (https://companyscope.io/register/air-2026-003). Entry IDs are stable; corrections publish as dated addenda on this page.

This analysis is the work Janus Compliance does for clients before the incident. For ongoing agent governance support, see Janus DPO-as-a-Service. Browse the full register or the vendor compliance index.