A tribunal held Air Canada liable for its chatbot's answer

A bereavement fare and a wrong answer

In November 2022, Jake Moffatt asked a support chatbot on Air Canada's website about bereavement fares after a family death. The decision, issued by the British Columbia Civil Resolution Tribunal on 14 February 2024 as Moffatt v. Air Canada, records that the chatbot told Moffatt they could "submit your ticket for a reduced bereavement rate" within 90 days of booking, even retroactively. A separate Air Canada webpage, linked from the chatbot's own answer, stated the opposite: bereavement consideration did not apply after travel was completed. Moffatt booked flights relying on the chatbot's statement and was later refused the reduced fare.

What the tribunal rejected

Air Canada argued, in the tribunal's words, that it "cannot be held liable for information provided by...a chatbot", in effect treating the tool as a separate legal actor responsible for its own statements. Tribunal member Christopher Rivers called this "a remarkable submission", writing that "it should be obvious to Air Canada that it is responsible for all the information on its website" and that "it makes no difference whether the information comes from a static page or a chatbot". The decision also rejected Air Canada's unsupported claim that its tariff excused the error, noting the airline never produced the tariff text it relied on, and it drew an adverse inference from Air Canada's failure to provide evidence of the actual bereavement fare.

A narrow claim, a broad implication

The tribunal found negligent misrepresentation on ordinary consumer-law grounds - a duty of care, an inaccurate statement, reasonable reliance and resulting loss - not a novel rule written for AI. It ordered Air Canada to pay CAD 812.02, combining damages, pre-judgment interest and tribunal fees. That is a small-claims outcome, not a data point about chatbot accuracy generally; it establishes only that a company cannot disclaim responsibility for its own customer-facing tool by pointing at the tool. OWASP's guidance on misinformation risk in LLM applications later cited this outcome as a real-world example of overreliance on an unverified chatbot answer.

• If a chatbot and a static page on the same site could give conflicting answers, which one governs, and does anything on the site say so?
• What is the process for correcting a chatbot's answer once a discrepancy with official policy is identified?
• Is there a record, of the kind a tribunal would accept, of what a chatbot actually told a specific customer?

The decision does not say a chatbot must never be wrong; it says a business does not get to treat its own deployed assistant as an entity separate from itself when the assistant is wrong. That distinction, more than the modest damages, is the reason the case is cited well beyond small-claims practice.