As AI models become increasingly capable of performing tasks traditionally reserved for humans, such as decision-making, creative work, and even ethical reasoning, to what extent should we trust AI systems to make critical decisions, and how can we ensure accountability in cases of failure or harm?