Turning the Hiroshima AI Process into Rules That Companies Can Use

The Hiroshima AI Process showed the world a clear direction: the goal is not simply to make AI larger, but to deploy it across society in ways that are safe, secure, and trustworthy.

It was the first international policy framework for advanced AI systems to include both International Guiding Principles and an International Code of Conduct.

The next step is to turn those principles into rules that can operate in real business environments.

▼ Read the press release(JP) https://prtimes.jp/main/html/rd/p/000000005.000182721.html

The Gap Between Principles and Formal Adoption

Use AI safely.

Preserve the basis for each decision.

Ensure that decisions can be explained when something goes wrong.

These are sound principles. But principles alone are not enough for a company to formally adopt an AI-generated decision in production.

Businesses need more concrete rules:

• What must be verified before a decision can be accepted?

• Where must the process stop if the required evidence is missing?

• How can the company confirm that the conditions or records were not changed after the decision?

• If a problem occurs, can the company verify why the decision was allowed to proceed?

A significant implementation gap remains between international principles and the concrete conditions companies need to use AI in real operations.

From Joint Implementation with ONZALINX to Common Rules

GhostDrift Mathematical Institute and ONZALINX have been developing this implementation layer, beginning with pharmaceutical cold-chain logistics.

The first phase of the machine-verification PoC marks the first milestone in that effort.

Its significance goes beyond the completion of a single logistics PoC.

• Connect operational conditions with the evidence required to verify them.

• Allow only decisions that may legitimately proceed.

• Stop the process when the predefined conditions are not satisfied.

• Make it possible to verify afterward why a decision was accepted or stopped.

The value of the joint implementation lies in turning this structure into something that can actually operate in the field.

The next stage begins here.

We will extend the structure implemented in pharmaceutical logistics beyond that specific operation and develop it into common rules that can be used across different companies and business processes, including quality verification, receipt approval, inventory confirmation, and authorization checks.

Hiroshima AI Assurance Protocol

To support that expansion, we are laying the groundwork for the development of the:

Hiroshima AI Assurance Protocol

This initiative will be advanced through the Hiroshima AI Assurance Council, which is currently being established.

GhostDrift Mathematical Institute is one of the founding initiators, while ONZALINX—headquartered in Hiroshima—will serve as the council’s representative.

This is not an official initiative of the Hiroshima AI Process.

It is an independent, private-sector effort that carries its principles forward by translating them into concrete rules that companies can use to formally adopt AI-generated decisions in production.

The protocol will address information such as:

• What the AI observed

• Which conditions it verified

• What evidence supported its decision

• Why the decision was allowed to proceed

• What state change resulted from the decision

The goal is to make this information verifiable across companies and systems.

An AI-generated decision should not remain merely a recommendation.

It should become a formal decision that a company can adopt responsibly.

We will develop the common foundation needed to make that possible, beginning with real-world implementation in Hiroshima.

Image: Turning the Hiroshima AI Process into Real-World Implementation

Turning Japan’s Operational Strength into Rules the World Can Use

AI competition will not be determined by model size and generative performance alone.

Another part of the competition will be over who defines the common rules for determining what makes an AI decision trustworthy—and under what conditions society may formally accept it.

For decades, people working across Japan’s industrial front lines have carefully defined conditions, maintained records, and worked to ensure that anomalies are not overlooked.

AI does not make that work obsolete.

As AI begins to make decisions and act in the real world, that accumulated knowledge can become the foundation for teaching AI what it must verify and when it must stop.

We will translate this operational knowledge into rules that AI can interpret and machines can verify.

We will extend processes used within a single company into structures that can operate across organizations—and develop implementations created in Japan into protocols that can be used around the world.

The commitment to quality that people across Japan’s industrial front lines have sustained for decades is now taking on a new meaning: it can become a foundation for using AI responsibly across society.

The joint PoC in pharmaceutical logistics is the first milestone.

From a single logistics decision to a common foundation that can operate across companies, industries, systems, and national borders.

From the Hiroshima AI Process to real-world implementation.

From Hiroshima, we will develop a Japan-originated AI assurance protocol for the world.