Yeah the Air Canada case probably isn't a big indicator on where the legal system will end up on this. The guy was entitled to some money if he submitted the request on time, but the reason he didn't was because the chatbot gave the wrong information. It's the kind of case that shouldn't have gotten to a courtroom, because come on, you're supposed to give him the money any it's just some paperwork screwup caused by your chatbot that created this whole problem.

In terms of someone someone getting sick because they put glue on their pizza because google's AI told them to... we'll have to see. They may do the thing where "a reasonable person should know that the things an AI says isn't always fact" which will probably hold water if google keeps a disclaimer on their AI generated results.