21.5 AI Safety

AI safety is about making sure AI systems behave reliably and do not cause harm, even in unexpected situations. It includes alignment (the system pursues the goals we actually intend), robustness (it handles unusual or adversarial inputs gracefully), and guardrails (limits that prevent dangerous outputs).

For example, a customer-service chatbot for an Ethiopian bank must refuse to reveal another customer's balance, resist 'jailbreak' prompts that try to trick it, and escalate to a human for anything it is unsure about. These are safety guardrails, not just polite behavior.