We describe our latest thinking in the hope of helping other AI developers address safety and misuse of deployed models....