The fix, imho, is for the reviewers to also use ai to review the code. However, the ultimate responsibility for the outcome(s) should be on the committer - you commit it, you own it, so to speak. If there's an incident, they need to be the one paged in the middle of the night. Bugs resulting from it will land on their desk.
The reviewers aren't a shield/safety net.
