Expert: V.S. Subrahmanian, Professor of Computer Science, Northwestern University The error here is a single sentence, and it is the kind that costs a customer money. United's chatbot told Alison Gil that her $200 TravelBank credit would "stay active for 5 years from the date it's deposited." United's actual rule, as the airline confirmed to her and later to NBC 5 Responds, was one year - her credit was set to expire on September 27, 2026. The airline has acknowledged the bot was wrong: its "AI chatbot provided this customer with the wrong information for her situation." Two things are worth separating. The first is the factual error itself: an automated system stating a concrete expiry rule with a specific number of years, when the number did not apply to the customer in front of it. United says the expiration of TravelBank funds varies with how the funds were received, which is precisely the condition a customer cannot check for themselves - they asked the company's own assistant to tell them. Gil described the outcome bluntly: "There seems to be no accountability in it. Their chatbot can apparently give out whatever answer it wants." The second is what the airline's remedy consisted of. It apologised, told her the credit could not be extended, and only after a reporter contacted the airline did it issue a $200 electronic travel certificate. The standing disclaimer it points to - that chatbot information "may not always be complete or relevant" - is the mechanism by which a confidently wrong answer is pre-emptively excused rather than prevented. V.S. Subrahmanian, a professor of computer science at Northwestern University, puts the responsibility question plainly: "We know human operators will make mistakes. We expect AI to make mistakes as well, but there should also be some responsibility for those mistakes." On why these failures recur, he points to what the bot cannot do: "I think the one thing that chatbots lack, which many humans have, but not all, of course, is the ability to dynamically listen to what you're saying and tell if it sounds right or not." His conclusion is structural: "There needs to be some regulatory structure that has some system of checks and balances." Gil's case is not an isolated support failure. NBC 5 reports other customer complaints about airline chatbots, including a Chicago-area traveller who said a question about a refund fee led a United chatbot to cancel their entire trip. The article also notes that while no study has specifically measured AI customer-service chatbot accuracy, at least two studies of major AI chatbots found that roughly half of responses contained errors, misleading information or missing context. The practical takeaway consumer advocates give is the one Gil's screenshot made possible: save the transcript. Her case moved only because she could show the exact sentence the bot produced - and because a journalist asked the airline about it. Source: https://www.nbcchicago.com/consumer/ai-chatbots-are-taking-over-customer-service-but-how-often-are-they-wrong/3989958/