RPO specifies the maximum acceptable data loss measured in time (e.g., "no more than 4 hours of data"); RTO specifies the maximum acceptable downtime before service restoration. These metrics are standard in disaster recovery planning but require special attention for AI systems. Contracts should specify RPO/RTO for each critical component (models, vector stores, knowledge bases, configuration) and clarify whether RTO includes re-indexing and validation time, not just data restoration.
See: Availability; Business continuity; Service Level Agreement / Service Level Objective; Vector database