In addition, AI systems should integrate safety and security by design mechanisms to ensure that they are verifiably safe at every step, taking at heart the physical and mental safety of all concerned.