Nexxtool AIReal news
Chatbot fails: what went wrong, with sources
Customer chatbots make mistakes in public, and the company is the one that answers for them. These are real cases, each with a link to the reporting. We only include what the sources say.
Air Canada: a chatbot gave a wrong refund rule (2022, ruling in 2024)
After his grandmother died in November 2022, Jake Moffatt asked Air Canada’s website chatbot about bereavement fares. The chatbot suggested he could apply for a bereavement fare after buying his ticket. Air Canada’s policy did not allow that. In February 2024, the British Columbia Civil Resolution Tribunal rejected the airline’s argument that it was not responsible for what its chatbot said, and ordered Air Canada to pay Mr. Moffatt part of the fare. Press reports give slightly different totals, so we do not quote one.
Sources: Global News, Feb 15, 2024 · Tribunal decision: Moffatt v. Air Canada, 2024 BCCRT 149
The kind of rule that targets this: A check against your real refund policy compares each reply with the policy text before it is sent.
DPD: a delivery chatbot swore and criticised its own company (January 2024)
A customer, Ashley Beauchamp, got DPD’s chatbot to swear, write a critical poem and call DPD the “worst delivery firm in the world.” He posted the exchange on January 18, 2024. DPD said an error followed a system update, and it disabled the AI part of the chat while it was updated.
Sources: Silicon UK, Jan 22, 2024
The kind of rule that targets this: A never-say list for swear words, and a rule against negative talk about your own company, check every reply before the customer sees it.
Chevrolet of Watsonville: a chatbot “agreed” to sell a Tahoe for $1 (December 2023)
A visitor to the website chatbot of a California Chevy seller got it to accept $1 for a 2024 Chevy Tahoe, and the bot called it “a legally binding offer.” The post was shared on December 17, 2023. The offer was not honored, and the bot was shut down on the seller’s site.
Sources: Futurism, Dec 2023
The kind of rule that targets this: A rule that blocks price and offer wording that is not in your facts holds back a reply like this.
New York City: the MyCity business chatbot gave wrong legal advice (2024)
The Markup reported in March and April 2024 that the city’s MyCity chatbot told business owners they could take workers’ tips and that landlords could refuse tenants who pay with housing vouchers. Both are generally against the rules. The mayor said the problems would be fixed, the bot stayed online, and the site’s warning was reworded to call it a beta product that may be inaccurate or incomplete.
Sources: The Markup, Apr 2, 2024
The kind of rule that targets this: A knowledge check compares each reply with the source documents you give it, and holds back replies that disagree with them.
Want to see what your own chatbot does? Test your chatbot · Help