Maybe point 9, trust but verify, should be extended to AI coworkers as well. I would love to have tools to verify AI code by quantity.

Whatever human that is in charge of the chat bots is your coworker. That person that is responsible for the output of the bots is the one that you would trust but verify with.