That is scary but the methods traditionally used to deny claims aren't really any better. I've had claims denied after they were explicitly pre-approved because of string literals not matching exactly.

It at the very least provides more cover to the ones denying the claims. They can blame it on AI in the hopes they're not the next one being targeted by vigilantes.

My aunt worked for an insurance company while she was semi-retiring as a doc, she lasted a few months before she was too disgusted to continue.

AI isn't needed for insurance to fuck anyone over.