Levels of Concern

the paradox is that once you're advanced enough to know what security concerns are worth your time, you probably have already internalized this.

I think the devs who are dutifully reviewing the AI for pixel-pushing tasks simply do not trust the AI at all, and would probably be happier and just as productive not using it. you need a baseline level of trust to use the tool effectively.