The Spread of Physical AI and Governance Challenges

The advancement of Physical AI, which is expanding into robots, sensors, and industrial equipment, is making governance issues for autonomous systems increasingly important. Beyond simply whether AI agents complete tasks, the key issue is how to test, monitor, and halt actions that occur in real-world environments.

Market Size and Outlook

• Industrial Robots: Global installations reached 542,000 units in 2024, more than double the figure from a decade ago, and are projected to exceed 700,000 units by 2028.
• Physical AI Market: Expected to grow from approximately $81.6 billion in 2025 to $960.3 billion by 2033.

What Sets Governance Apart Unlike software-only AI, Physical AI interacts directly with workspaces, infrastructure, and human users. Since model outputs translate directly into robot movements or machine commands, safety limits and escalation paths must be built into the system design from the outset.

Google DeepMind's Case Google DeepMind has introduced Gemini Robotics and Gemini Robotics-ER for robotics.

• Gemini Robotics: A vision-language-action (VLA) model that directly controls robots.
• Gemini Robotics-ER: Focuses on Embodied Reasoning, including spatial understanding and task planning.

For Physical AI to succeed, it requires Generality to handle unfamiliar environments, Interactivity to respond to human input, and Dexterity to achieve precise movements.