Frontier Model Evaluations

Assessing the most capable models for dangerous capabilities.

Why are 'frontier model' evaluations especially emphasized in AI governance?