Safety Evaluation Enters the Foundation-Model Stack

Google paired the robotics release with layered safeguards and an agentic-robot safety benchmark.

DXresearch interpretation: AI capability must remain bounded by independent safety controllers and verified sensing chains.