On 14 February 2024, British Columbia's Civil Resolution Tribunal ruled in Moffatt v. Air Canada that the airline was liable for a wrong answer its website chatbot gave a grieving customer. On 11 November 2022, the day his grandmother died, Jake Moffatt asked the chatbot about bereavement fares. It told him he could buy a full-price ticket and claim the reduced rate within 90 days. Air Canada's policy did not allow bereavement fares to be claimed retroactively, so it refused the refund. Before the tribunal, Air Canada argued it could not be held liable for information given by its agents or representatives, including a chatbot. Tribunal member Christopher C. Rivers rejected that, found negligent misrepresentation and ordered Air Canada to pay $812.02 (Canadian dollars). No system was breached. The case is on this list because an AI agent acted for an organisation, gave out information nobody had approved, and the organisation was held responsible for what it said.
Key takeaways
- Air Canada's website chatbot told a customer he could apply for a bereavement fare within 90 days of buying his ticket, which contradicted the airline's published policy, according to the tribunal decision.
- Air Canada argued it could not be held liable for information provided by its agents, servants or representatives, including a chatbot. The tribunal called this "a remarkable submission" and rejected it.
- The tribunal ordered Air Canada to pay $812.02: $650.88 in damages, $36.14 in interest and $125 in fees. The sum is small, but the principle was widely reported.
- This was an AI-agent incident, not a security breach: no credential was stolen and no attacker was involved. The chatbot gave a wrong answer on its own in an ordinary conversation.
- The identity lesson: an AI agent that speaks for an organisation acts with the organisation's authority, so it needs an owner, limits on what it may commit to and a record of what it said.
At a glance
| Organisation | Air Canada |
|---|---|
| When | Chatbot conversation on 11 November 2022; tribunal decision issued and made public on 14 February 2024 |
| Attacker | None. The chatbot's wrong answer surfaced when the customer, Jake Moffatt, claimed his refund and then took Air Canada to the tribunal |
| Entry point | An ordinary customer question to the chatbot on Air Canada's website about bereavement fares |
| Identities abused | A customer-facing AI chatbot speaking for Air Canada; the model and vendor behind it were not disclosed |
| Impact | Air Canada was found liable for negligent misrepresentation and ordered to pay $812.02; no data or system compromise |
| Category | Agentic AI and AI agents. Incident class: AI-agent incident (a chatbot gave unapproved policy information that the operator was held liable for) |
What happened
Jake Moffatt's grandmother died in Ontario on 11 November 2022. The same day he visited Air Canada's website to book a flight and used the chatbot to ask about bereavement fares. According to the decision, the chatbot replied that Air Canada offers reduced bereavement fares and that a customer who had already travelled and wanted a reduced rate should "kindly do so within 90 days of the date your ticket was issued" by completing a refund application form. The answer included a link to Air Canada's bereavement travel page, which, MobileSyrup reports, said the bereavement policy does not apply retroactively.
Relying on the chatbot, Moffatt booked a one-way flight from Vancouver to Toronto for 12 November at $794.98 and, on 16 November, a return for 18 November at $845.38, a total of $1,630.36. He first applied for the partial refund on 17 November 2022. In February 2023 he sent Air Canada a screenshot of the chatbot's answer. An Air Canada representative replied on 8 February 2023, admitting the chatbot had provided "misleading words", but still refused the refund. Air Canada later offered a $200 coupon, which Moffatt did not accept. He claimed $880 at the Civil Resolution Tribunal, the body that hears small claims in British Columbia.
Air Canada's main defence was that it was not responsible for the chatbot. "Air Canada argues it cannot be held liable for information provided by one of its agents, servants, or representatives," the decision records, including a chatbot. Tribunal member Christopher C. Rivers wrote: "In effect, Air Canada suggests the chatbot is a separate legal entity that is responsible for its own actions. This is a remarkable submission." He added: "While a chatbot has an interactive component, it is still just a part of Air Canada's website. It should be obvious to Air Canada that it is responsible for all the information on its website. It makes no difference whether the information comes from a static page or a chatbot."
He also rejected the argument that the customer should have checked the linked policy page, noting that Air Canada did not explain "why customers should have to double-check information found in one part of its website on another part of its website." He concluded: "I find Air Canada did not take reasonable care to ensure its chatbot was accurate." The tribunal awarded $650.88 in damages, plus $36.14 in pre-judgment interest and $125 in tribunal fees. Commentary in the American Bar Association's Business Law Today summed up the lesson as "companies remain liable for the actions of their AI tools".
Timeline
| Date | Event |
|---|---|
| 11 November 2022 | Moffatt's grandmother dies; he asks Air Canada's chatbot about bereavement fares and books a flight to Toronto. |
| 16 November 2022 | Moffatt books his return flight to Vancouver. |
| 17 November 2022 | Moffatt first applies for the partial bereavement refund. |
| 8 February 2023 | An Air Canada representative admits the chatbot gave "misleading words" but refuses the refund. |
| 14 February 2024 | The Civil Resolution Tribunal issues its decision and orders Air Canada to pay $812.02. |
| 16 February 2024 | The ruling is widely reported, including by MobileSyrup. |
How it happened: the identity attack path
- Agent deployed to speak for the company. Air Canada put a chatbot on its website to answer customer questions, giving it the same standing as any other page the airline published.
- Unverified answer on a policy question. Asked about bereavement fares, the chatbot produced a rule that contradicted the published policy, telling the customer he could claim the reduced fare within 90 days of ticketing.
- Customer relies on the agent. Moffatt booked full-price flights on the strength of that answer, as any customer reading the airline's website would.
- Operator disowns the agent. Air Canada refused the refund and argued it was not liable for what its chatbot said, treating the agent as if it were outside the organisation.
- Accountability assigned. The tribunal held that the chatbot was part of Air Canada's website and that the airline had not taken reasonable care to make it accurate.
Impact
- Confirmed: Air Canada was found liable for negligent misrepresentation and ordered to pay $812.02 within 14 days.
- Confirmed: no security compromise; no data was exposed and no credential was misused.
- Wider: the decision became a widely cited example that an organisation cannot treat its AI chatbot as a separate party. It is a small claims decision in one Canadian province, so its legal weight elsewhere is limited, but the reasoning is simple and has been repeated in legal commentary.
- Potential: any organisation whose agent makes commitments on price, refunds or policy faces the same exposure, multiplied by the number of conversations the agent has.
What this means for NHI and AI agent security
This case has nothing to do with stolen keys, yet it belongs in a non-human identity database because it settled a basic question of agent identity: whose authority does an AI agent act under? Air Canada tried to treat its chatbot as something apart from the company. The tribunal said the opposite. An agent deployed on an organisation's channel speaks with that organisation's voice, and the organisation answers for it. That is the same principle identity teams apply to service accounts and API keys: the actions of a non-human identity are attributed to whoever owns it.
The practical lesson is governance before deployment. A customer-facing agent needs a named owner, a defined scope of topics it may answer, a rule for questions it must hand to a human or answer only from approved text, and logs of what it said. The same discipline applies when agents gain the ability to act, not just answer, such as issuing refunds or changing bookings. Our Agentic AI Compliance Guide covers accountability for agent output, and the AI Agent Observability and Incident Response Guide covers keeping the records that let you see what an agent told a customer.
Recommendations
- Give every customer-facing agent a named owner. Someone in the business, not only the vendor, must answer for what the agent says and approve changes to it. See our NHI Ownership Guide.
- Restrict policy answers to approved sources. For prices, refunds, eligibility and legal terms, have the agent quote or link the current policy text rather than generate its own summary, and test those answers against the policy before release.
- Define what the agent may commit to. Treat promises about money or eligibility as an authorisation decision, and route them to a human or a rules engine. See our AI Agent Authorisation Guide.
- Log every conversation and keep it. In this case the evidence of what the chatbot said was the customer's own screenshot. Keep transcripts so you can find and correct wrong answers yourself. See our AI Agent Observability and Incident Response Guide.
- Honour the agent's answers when they are wrong, then fix the agent. Air Canada's attempt to disown its chatbot failed and became the most quoted part of the case. Build a complaints path that treats a wrong answer as an incident to correct.
- Review agents against the regulations that apply to you. Consumer protection and misrepresentation rules apply to automated channels too. See our Agentic AI Compliance Guide.
Frequently asked questions
What did the Air Canada chatbot get wrong?
On 11 November 2022 it told Jake Moffatt he could buy a full-price ticket and apply for a bereavement fare within 90 days of the ticket being issued. Air Canada's actual policy did not allow bereavement fares to be claimed after travel, so the airline refused his refund.
Was Air Canada held responsible for its chatbot?
Yes. In Moffatt v. Air Canada, decided on 14 February 2024, the British Columbia Civil Resolution Tribunal rejected Air Canada's argument that it was not liable for information from its chatbot, found negligent misrepresentation and ordered it to pay $812.02 in damages, interest and fees.
Was the Air Canada chatbot hacked?
No. There is no evidence of any attack, prompt manipulation or data breach. The chatbot gave an inaccurate answer in a normal conversation, and the case was about who is responsible for that answer.
Related NHI Mgmt Group resources
DPD Chatbot Incident 2024 · Chevrolet Dealer Chatbot 2023 · Replit AI Agent Database Deletion 2025 · Agentic AI Compliance Guide · AI Agent Authorisation Guide
How NHI Mgmt Group can help
Organisations are deploying AI agents faster than they are deciding who owns them and what they are allowed to say or do. We help teams put ownership, scope and logging around agents before they reach customers. See our NHI and AI agent security training.
References
- BC Civil Resolution Tribunal: Moffatt v. Air Canada, 2024 BCCRT 149 (14 February 2024)
- MobileSyrup: Air Canada has to pay for mistake made by chatbot, B.C. Civil Resolution Tribunal rules (16 February 2024)
- AI Business: Air Canada Held Responsible for Chatbot's Hallucinations (20 February 2024)
- ABA Business Law Today: BC Tribunal Confirms Companies Remain Liable for Information Provided by AI Chatbot (February 2024)