AI Safety Foundation Established to Support Safe Human-AI Collaboration — Comprehensive Guidelines and Evaluation/Verification Infrastructure for Design, Assessment, and Operation —

In the 'Project for the Promotion of R&D and Verification for AI Safety Assurance / R&D for AI Safety Enhancement' promoted by NEDO, five parties—NEDO, the National Institute of Advanced Industrial Science and Technology (AIST), Citadel AI Inc., Corpy & Co., Inc., and the University of the Ryukyus—have developed, formulated, and released guidelines and evaluation protocols that will serve as a common foundation for ensuring the safety of AI systems.
The guidelines formulated in this project focus on ensuring safety from the planning and design stages to the evaluation and operation of AI systems. They organize the concepts and basic procedures for businesses developing and implementing AI-powered systems to identify risks and consider appropriate countermeasures.
By widely promoting the use of these guidelines and evaluation protocols, the project aims to instill a common understanding and procedure for AI system safety in society, thereby accelerating the development of a common AI safety foundation for the secure use of AI.

1. Background
Against the backdrop of initiatives like the Hiroshima AI Process launched at the G7 Hiroshima Summit in 2023, discussions and system development concerning AI safety are advancing in various countries. In light of these international trends, Japan has also established the AI Safety Institute (AISI) to participate in international discussions.
This project supports public-private efforts from an R&D perspective to promote international rule-making for the safe and secure use of generative AI within Japan. As instances of humans and AI collaborating in decision-making and action increase, there is a growing need to address the common challenge of how to design, evaluate, and operate AI safety.
This project conducted R&D aimed at establishing a common foundation for evaluating and operating AI safety. While AI implementation technologies and application fields are diverse, the challenges of how to safely design human-AI interaction and how to ensure safety through judgment, verification, and operation are common across fields.
As shown in Figure 1, this project is structured to develop evaluation and management technologies that serve as a 'yardstick' for safety (R&D Item 1), develop AI safety evaluation and implementation technologies for specific application areas assuming real-world environments (R&D Item 2), and organize and systematize these results into a form usable by companies in practice, leading to the formulation of guidelines for AI safety implementation (R&D Item 3).
This project does not aim to demonstrate the completion of social implementation for any specific AI, but rather to present a common foundation that will lead to future technological development, demonstration, and standardization.

2. Achievements
In this project, a wide range of guidelines, evaluation methods, templates, and evaluation environments were developed, spanning the 'design, evaluation, and operation' stages to address the diverse challenges of AI safety.
(1) Formulation of Multimodal AI Quality Management Guidelines (R&D Item 3)
AIST, as a core achievement of this project, formulated guidelines that organize the perspectives and processes of quality management for multimodal AI, which receives images and text and responds primarily with text. Focusing on cross-modal correspondence capability as a unique basic evaluation perspective for multimodal AI, this capability is classified into four levels. To ensure the safety and quality of multimodal AI systems, it is particularly important to determine the required level of cross-modal correspondence capability, and the guidelines systematically organize the responses to be implemented at each stage of the lifecycle according to that level.
The guidelines also feature three case studies—automatic image captioning, image-based diagnosis of infrastructure deterioration, and content moderation on social media—to illustrate points to consider and quality management issues in situations involving human judgment and supervision.
These guidelines present a common design and evaluation framework for ensuring safety and quality based on the characteristics of multimodal AI, serving as a base for the practical application of AI safety.

(2) Formulation of Guidelines and Case Studies to Support AI Social Implementation in Corporate Settings (R&D Item 3)
Citadel AI, through interviews with companies actually developing and operating generative AI and AI agents, has organized evaluation perspectives, standards, and methods to translate AI safety principles and guidelines into a form that can be implemented and operated on-site. Common patterns and practical know-how were extracted from technological, process, and organizational culture perspectives and systematized into a 'Generative AI Practical Guide and Corporate Casebook'.
To verify and concretize the extracted know-how for developers, a chatbot was implemented and released. It is designed for users who require special consideration in understanding language and systems, such as foreigners, and focuses on information related to administrative procedures. The chatbot utilizes information published by local governments to clearly present the generative AI's answers and their sources.
These achievements present the 'ideal state' shown in the guidelines as concrete evaluation and operation methods for companies to implement in practice, playing a role in making AI safety applicable at a practical level.

(3) Development of Implementation Guides and Evaluation Templates Connecting Organizational Management and Technical Evaluation (R&D Item 3)
Corpy has developed an 'AI Management System-based Generative AI Safety Evaluation Protocol and its Implementation Guide' and evaluation templates, consisting of three phases—analysis, testing, and reporting—for the practical implementation of generative AI safety evaluations aligned with ISO/IEC 42001 (AI Management System Standard).
Using a customer support system with a visual language model as a case study, evaluations including red-teaming were conducted to examine their effectiveness and practical issues.
These achievements provide a practical framework connecting organizational management requirements with technical safety evaluations, forming the basis for companies to consistently implement AI safety.

The following two points are achievements that lead to concrete ideas and evaluation/verification methods for ensuring safety in human-AI collaboration, assuming AI utilization in real-world environments such as healthcare and daily life.
(4) Proposal of a Method for Ensuring Safety in the Decision-Making Process in Human-AI Teaming (R&D Item 1)
AIST and the University of the Ryukyus jointly conducted a study on safety in Human-AI Teaming, where humans and AI collaborate on judgments, assuming a medical setting. This was an effort to concretize the idea of 'quality requirements according to use and context' presented in the multimodal AI guidelines.
Specifically, they analyzed what process should be used for final decision-making when AI and a doctor's judgments in medical image diagnosis do not match, and organized methods for consensus-building and risk avoidance.
Through joint discussions with doctors, they organized and systematized design principles for ensuring safety in the human-AI collaborative decision-making process, covering what additional information and judgment rationale AI should present, and how the interface should support decision-making.
They also confirmed that in relationships where humans have the final decision-making authority, humans may not be able to correctly perceive the AI's accuracy, and that this perception can affect the utilization of the AI.
This achievement organizes the concept of ensuring safety on the premise of situations where human and AI judgments do not match, and will contribute to the practical design of AI safety in Human-AI Teaming.

(5) Construction of an AI Safety Evaluation and Verification Platform for Daily Life Domains (R&D Item 2)
AIST has worked on building an evaluation and verification platform for AI safety to promote the safe use of AI systems in daily life environments such as homes and nursing facilities.
AI systems for daily life face challenges such as the difficulty of acquiring data for development and verification due to the diversity of living situations and privacy protection concerns, as well as the lack of well-established evaluation methods for safety and robustness.
In this initiative, targeting monitoring AI systems, they organized anticipated behaviors and events from a human safety perspective into scenarios and established a real-virtual fusion environment that enables the collection, generation, and verification of data necessary for AI safety evaluation.
By combining actual behavior data acquired in physical living labs built in the real world with data augmentation technology in a cyber living lab in a virtual space, they have constructed a dataset that contributes to the safety evaluation of monitoring AI, including behaviors that are difficult to measure in reality, such as falls and stumbles.
Furthermore, looking ahead to the social implementation of monitoring AI, the 'Human-centric AI Lifetech Consortium (HAIL)' was launched in April 2026 and will begin activities in June, as a forum for discussing technical and social issues related to safety evaluation and verification. This dataset is scheduled to be used in the consortium.
This achievement establishes a platform that makes it possible to verify the safety of AI in the daily life domain, which has been difficult to evaluate, based on the characteristics of the real environment, and will contribute to the advancement of practical evaluation and verification of AI safety.

3. Future Plans
This project has produced a variety of results, including the core formulation of AI safety guidelines, as well as implementation methodologies for corporate settings, practical methods linking organizational management and evaluation, technical knowledge on human-AI collaboration, and verification platforms for real-world environments.
These results are positioned as elements for a multifaceted approach to the diverse challenges of AI safety at each stage of 'design, evaluation, and operation,' and are characterized as a stepping stone to broadly realize countermeasures for expanding AI use cases.
In the future, we plan to systematically organize the common understanding of AI safety obtained in this project, along with specific evaluation and implementation methods, into a practical AI safety common foundation for a society where humans and AI collaborate.