Which of the following controls would MOST effectively mitigate worst-case service disruption scenarios affecting an AI-based application system?