Researchers Highlight Urgent Need for Better Safety Metrics in AI Agents

safety in agentic trajectories is SOO important and we need better measurements