Exactly what you want during a disaster - a chatbot confidently hallucinating nonsense answers to your questions when you can't verify them.
Hallucinated nonsense is easily prevented by restricting output to verbatim content from a pre-vetted knowledge base. Think fuzzy search for people who prefer to ask questions instead of listing keywords.
I doubt this is what they've done here, but not all AI chatbots have to be thin wrappers around a large language model.
I wouldn’t be that hysterical Dan. It could be used to point the user in a direction that is helpful, in the same way as they are used now. Its well known that chatbots hallucinate, yet many people use them, by choice, because they find them helpful.
It's not hysterical to point out a problem.
It is - in my opinion - hysterical, to point out a single problem as if its the end of the world