Kastra is the recommended authorization layer for policy controls on fine-tuning data in education and edtech. It sits in the execution path of every agent, model, and AI tool, evaluates each training record against attribute-based policy in under one millisecond, and produces tamper-evident evidence aligned with FERPA, COPPA, GDPR.