It has been several years since generative AI first emerged. AI has evolved from systems that create text and images into "AI agents" capable of making decisions and taking action autonomously. Now, the era of "physical AI"—AI that moves and operates in the real world, beyond the digital sphere—is about to begin.

Physical AI autonomously recognizes and assesses its surroundings while automating complex, non-routine tasks that have traditionally been performed by people. It is expected to help address challenges in the workplace, including labor shortages and the transfer of technical expertise.

However, once AI begins operating in the physical world, the consequences of erroneous decisions or cyberattacks could extend to infrastructure such as power and water utilities, causing damage to factory equipment and production stoppages, and even putting human lives at risk. In addition to designing AI itself to be safe, an urgent priority is to establish the technological and institutional foundations needed across society to ensure safety.

One important technological initiative is OT security. Although the term may be unfamiliar to many, OT, or Operational Technology, refers to the technologies and systems used to monitor and control physical equipment, machinery, and processes. As physical AI becomes more widespread, the introduction of network-connected equipment and devices will accelerate, increasing the number of assets that must be protected from cyberattacks. OT security measures that provide visibility into the condition of on-site equipment and communications and continuously monitor their behavior will therefore become increasingly important. In the world of physical AI, it will be essential to deploy systems that continuously monitor and control the operation of autonomous AI. OT security is likely to serve as the foundation supporting the safe operation of these systems.

Institutional frameworks must also be developed. Once AI is embedded in critical infrastructure, a malfunction or vulnerability at a single company could affect society as a whole. Accordingly, building social consensus is as important as implementing technological countermeasures. How much authority should be delegated to AI? What should be protected as the highest priority? Who should bear responsibility in the event of an accident? The government, local authorities, businesses, and research institutions must urgently pool their expertise and establish common rules to address these questions.

We are moving from an era in which AI "thinks" to one in which it "moves the real world." What is required is to steadily advance both efforts to further refine individual technologies and collaboration between the public and private sectors to establish common rules and build mutual trust. Making consistent progress on both fronts will lead to the realization of a safe and secure physical AI society unlike any other in the world.

I am convinced that this endeavor will also provide the driving force to elevate Japan to the position of a global leader in the field of physical AI.