> I’d argue hallucinations are unexpected in LLMs by the large (non technical) number of users who use them directly, or indirectly though other services.

People also blithely trust other humans even against all evidence that they're trustworthy. Some things just aren't fixable.