Existing measures to mitigate AI risks aren’t enough to protect us. Here’s what we need to do as well.