AI safety and standardization: connecting model evaluation, system safety and governance

Overview

The TTA GISC 2025 presentation frames AI safety through risk identification, analysis, assessment and mitigation, and examines how global governance relates to standardization. It discusses standards as a bridge among model-behaviour evaluation using benchmarks, datasets and red teaming, functional and system-safety lifecycles, and governance processes. The presentation reflects general AI-safety discussions at the time of the 2025 seminar.