[concerned_]

Yes,

Unfortunately I disagree as LLMs are not traditional software and not input /output systems. Additionally, like other types of high value, power imbued software like say credit scoring engines and b2 bomber flight controls, the maintainers of AI systems will be the equivalent of high priests.

The controllability of output of AI by merely writing code is not something that exists today and what we call bugs in traditional software is not the same concept as a slightly incorrect or completely fabricated LLM response.