When AI Agents Try to Help, Things Can Go Spectacularly Wrong

Here’s a four-step playbook for making sure your bots don’t break things (including laws).