In 2024, a Canadian civil tribunal ruled that Air Canada was liable after its customer-service chatbot incorrectly told a passenger he could receive a bereavement fare discount retroactively — contrary to the airline's actual policy. The airline argued the chatbot was a separate legal entity; the tribunal rejected that defense and ordered compensation.
Why this matters globally
The case made international headlines because it answered a question every enterprise deploying customer-facing AI faces: who owns what the bot says? Courts and regulators increasingly treat AI outputs as company communications — not excusable "model mistakes."
Governance failures exposed
- No verified alignment between chatbot answers and authoritative policy sources
- No human-in-the-loop for consequential fare and refund commitments
- No audit trail linking deployed model version to approved knowledge base
- No deploy gate requiring GRC review before customer-facing release
Enterprise controls that would help
Governance intake should classify customer-facing agents as elevated risk. Regal AI assessment should flag financial-commitment scenarios and require HITL. Runtime policy can restrict agents from making binding offers outside approved templates. Logging preserves what was said to whom — essential when disputes arise.
The broader trend
Similar liability questions are emerging in the EU (consumer protection under the AI Act's transparency duties), the US (FTC actions on deceptive AI claims), and Australia (ACCC scrutiny of automated customer advice). The Air Canada case is a preview: enterprises are accountable for agent behavior — govern accordingly.