Technical

safety and controllability

Ensuring AI is transparent, robust, and traceable.

Definition

A composite concept emphasizing that AI systems should be transparent and interpretable, robust and reliable against failures or adversarial conditions, and traceable through audit trails and recordkeeping to enable oversight and control throughout the lifecycle.