AI safety is the discipline focused on preventing AI systems from causing harm, whether through accidents, misuse, or unintended behavior. It spans practical concerns, like ensuring a medical chatbot never gives dangerous advice or a bank's model cannot be tricked into approving fraud, up to long-term questions about controlling very capable systems. For businesses, safety translates into concrete practices: testing before deployment, guardrails on outputs, human oversight for sensitive decisions, and clear procedures when the AI gets something wrong.