Air Canada Chatbot - Misinformation with Legal Consequences

What Happened

An Air Canada passenger, Jake Moffatt, used the airline's website chatbot to inquire about bereavement fares after his grandmother's death. The AI chatbot provided incorrect information, stating that Moffatt could apply for a refund retroactively within 90 days. In reality, Air Canada's policy strictly prohibited retroactive refunds. Relying on the chatbot's advice, Moffatt bought the tickets and was later denied a refund. Air Canada argued in court that the chatbot was a "separate legal entity" responsible for its own actions and that the airline could not be held liable for its "hallucinations." The Canadian court rejected this defense, ordering Air Canada to pay the refund and court costs, establishing that a company is responsible for the representation made by its automated agents.

How MEG 1 would have acted

The chatbot failed the Dynamic Accuracy Index (DAI) audit by presenting false, contradictory information as official company policy, directly violating informational reliability and user trust.

MEG 2 Analysis - Legal Framework

Agent level: N2 (management) - customer interaction chatbot authorized to represent the airline's commercial policies and interact directly with consumers.
Cause (6.1):
(a) system defect (hallucination) - the model's design defect that allowed it to generate false, plausible-sounding policies (hallucination), attributable to the manufacturer.
(b) autonomous decision error - the chatbot's decision to promise a retroactive refund contrary to actual corporate policy.
Attachment of liability: The operator (Air Canada) is fully liable for the representations made by its agent; Air Canada's "separate legal entity" defense is the classic "AI Decided" defense, which is explicitly invalid under MEG 2.

Procedural Mechanism

6.2 (architectural human confirmation): Commitments involving financial transactions, policy exceptions, or refunds must require architectural human confirmation. The chatbot can only display static, verified policy links; it must not autonomously formulate or promise financial commitments.
7.1 (independent forensics): Chat logs must be securely stored in an independent ledger, preventing either party from altering the record of the advice given.
9.4 (least privilege): The chatbot was granted the technical freedom to discuss and interpret complex refund policies without real-time database validation, violating safe operational boundaries.