ANSWER
Human oversight means that a human must be involved in decisions made by an AI system, especially in high-risk applications. This is not a formality: humans must be able to understand the system, monitor it and intervene as needed. This is a core safeguard against automating unjust decisions.