I've seen AI experts say that it's entirely possible that a solution will NEVER be found to the hallucination/fabrication issue. Apparently it implicates a weakness which is at the very heart of LLMs.
If that's the case, seems like it would be wise to tone down the AI hype a good bit.