AI or no AI is the wrong metric, of course. I imagine the way to solve this usefully probably goes through something that looks like Reddit. More "meta-moderation."

The problem is "a number of submissions we weren't prepared for," not "AI."