[datpuz]

I would argue that they need heavy human oversight

[Cthulhu_]

For sure; I'll believe that an AI can read and "understand" code, extract meaning and requirements from it, but it won't be the same as a human that knows requirements.

Then again, a human won't know all requirements either; over time, requirements are literally encoded.

[abdullin]

In systems like that you can record human interactions with the old version, replay against the new one and compare outcomes.

Is there a delta? Debug and add a unit test to capture the bug. Then fix and move to the next delta.