The Failure Rate of Language Models

The failure rate of language models will hold them back from being used in the world to solve actual problems. Could there be similar problems where language models will confabulate or hallucinate on say untrue things and therefore they can't be used as productively as we wish they could? I think it's very likely in many domains, but different domains have different kinds of reliability. It might be that now we just use these models to send an enormous amount of verbiage that no human ever reads. But maybe that if that saves a human nerves and ice that might be a good idea.