I think if you look at #rustledger (just as an example) right now you'll see the oversight and quality is pretty high. Why shouldn't it stay that way ? Isn't this a counterexample to what you're saying ? My 2c: not all use of AI is the same; not all of it is vibe coding; and projects where AI is used in some way, don't have to be low quality.
This aside, I understand the wish to have your accounting be on a rock solid foundation - I feel the same. No software is perfect, but you can still achieve it: keep your data, and if you need extra assurance, your reports, in version control; use double entry bookkeeping; regularly reconcile with real world balances; and perhaps run extra correctness checks. (#plaintextaccounting FTW.)
It is now very hard to miss any mistake made by your software - you'll see and address it right away.
@simonmic I submitted some fixes to rustledger and the maintainer rehashed them through an LLM that botched them. I sent corrections that were either hand applied or more strongly hand-held. Then the next day they regressed again because claude didn't like my fixes and reverted them while austensibly working on something else and the maintainer didn't even notice the unrelated changes.