Toward AI safety and its trustworthy materialization AI Ethics Research
AI is expanding in various fields, and new technologies such as generative AI are being developed. On the other hand, the ethical problems caused by AI are coming to light. In order to realize AI that can be used with trust, Fujitsu will comply with laws and guidelines, and carry out R&D on AI that is acceptable to society and in line with social values and norms.
Latest information AI Ethics Research
- March 14, 2025 Yuri Nakao, Senior Research Manager of Fujitsu, presented at the GLOCOM Symposium "Japan's AI Governance and Role in the World". [Japanese Only]
- March 6, 2025 Izumi Nitta, Senior Research Manager of Fujitsu, spoke as a panelist at the International Women's Day Special Event: "Inclusion Through Innovation: Can AI Bridge the Gender Gap in Society?"
- March 4, 2025 Yuri Nakao, Senior Research Manager of Fujitsu, presented at the AIST symposium "4th Grand Canvas: Drawing the Future of AI Quality". [Japanese Only]
- February 18, 2025 A new video of technology insight "2025 The Year of AI Compliance Understanding the EU Act" has been released from Fujitsu Research of Europe.
- February 6, 2025 Virginia Ghiara, Senior Research Manager of Fujitsu, presented at the MyData Online Symposium: Human-centric AI between Innovation and Compliance.
- January 21, 2025 "A Statistical Analysis of LLMs' Self-Evaluation Using Proverbs" was presented at PRICAI 2024. [Japanese only]
- October 18, 2024 Gender x AI research in collaboration with Ochanomizu University was reported at the 2nd Fujitsu SRL conference. [Japanese Only]
- October 14, 2024 Virginia Ghiara, Senior Research Manager of Fujitsu, presented at a panel session in the Business at OECD (BIAC) event.
R&D Pillars
Fujitsu is committed to complying with laws and guidelines and to developing technologies to design, develop, and operate AI safely and securely. We will also take on the challenge of solving social issues through collaboration between humans and AI, aiming to realize a more sustainable society.
Compliance and Guidelines
The development of AI and its utilization has been legislated and required to be observed. The EU has adopted the AI Act, and legal compliance in the development and use of AI will be mandatory in the EU by 2026.
Action: Partnership with international forums on AI governance
Fujitsu is one of the founding members of AI4People, the largest AI ethics organization in Europe. As an expert in this interdeciplinary community eligible in the impact of AI on society, we are promoting discussions on an ethical framework for creating a "Good AI Society". We contribute to four AI4People documents and co-chair scientific committees (2023-2024) to promote the ethical use of AI technology.
We also participate in the Organization for Economic Cooperation and Development, GPAI (The Global Partnership on Artificial Intelligence), and AI Safety Institute (AISI) to promote the ethical use of AI technology, and as a member of ISO/IEC for international standardization, CEN-CENELEC for European standardization, and BSI (British Standard Institution), we are engaged in activities that contribute to the benefit of society as a whole, such as establishing AI standards.
Ethics-by-Design for AI Governance
Fujitsu develops technologies to provide AI that conforms to social values and norms and is accepted by society.
Technology: AI Ethics Impact Assessment
This technology identifies and evaluates ethical risks of AI services before designing them. Based on AI ethics guidelines, you can systematically analyze the ethical risks that could be yielded by AI systems.
AI Ethics Practice Guide
Procedure manual for conducting AI ethics impact assessments and AI ethics models.
AI Ethics Impact Assessment Toolkit
Fujitsu AI Ethics Impact Assessment has been released on Fujitsu Kozuchi to reduce the manhours for AI's risk assessment. It is now open to the public. (The page is unavailable from 14:00 to 17:00pm (JST) every Thursday due to scheduled maintenance.
AI Ethics Impact Assessment White Paper
For developers and providers of AI systems, as well as customers using AI systems,
This document explains "AI Ethics Impact Assessment" for spreading trustworthy AI in society.
AI Ethics Impact Assessment Casebook
This document demonstrates outcomes of the AI Ethics Impact Assessment in AI use cases, providing an opportunity for developers, providers, and customers of AI systems to recognize that ethical issues can arise with AI systems.
Technology: Fujitsu AI Ethics for Fairness
Fujitsu AI Ethics for Fairness allows to verify fairness of AI models and data that is used to train them through web interface with a simple operation. It moreover improves fairness of the models even in the case where fairness across intersectional subgroups is essential. In addition, this technology is combined with Fujitsu AutoML to automatically build models optimized with respect to both fairness and predictive accuracy.
Technology: AI Risk Management Tool
It provides functionalities to manage ethical, quality, and security risks lurking through the entire AI lifecycle, and helps evidence that required measures for the risks took place to make things compliance with laws and requirements in guidelines.
Action: Open Source Software Projects
We hosted Intersectional Fairness and Sapient ML projects with Linux Foundation.
Action: Academic Collaborative Projects
University of Glasgow (2019～): we started a joint research on AI fairness with the University of Glasgow, with whom we had a long collaboration on personnel exchange. Following this, we continue to work closely with experts in Computer-Human Interaction on interdiciplinary research on participatory AI audit.
Technische Universität München (2022～2023): In a workshop-style dialogue with people in several fields, including finance, we jointly proposed an AI accountability framework that leads to industrial growth while adhering to the principles and guidelines of AI ethics.
Combating Biases in Generative AI
Generative AI has become a major concern of society. At the G7 summit held in Hiroshima in May 2023, there was discussion on international code of conduct for the development and use of generative AI. While generative AI is useful and beneficial, there are concerns that it may output misinformation or biased information. Therefore, we are developing technologies to make AI fair and safe to use.
Technology: Fujitsu LLM Bias Diagnosis
It diagnoses potential biases in LLMs from a variety of perspectives and suggests best LLMs for your business goal. Concretely, it evaluates and visualizes biases with which LLM could yield stereotypes on attributes (e.g. gender and age) and ethical problems rooted in societal and cultual differences.
White Paper
AI trust and Fujitsu's AI trust technologies in Conversational Generative AI - trust in AI (the trust that people and society have in AI) and AI trust (AI ethics, security, and quality) are critical to the spread of AI in businesses and society. The document describes Fujitsu's activities to achieve trust in AI and its AI trust technologies.
Challenges to Social Issues
As AI researchers, we are working on gender equality issues to achieve a better society for more people.
Action: Collaborative Research with Ochanomizu University
By combining AI ethics technology from Fujitsu and academic insighton gender equality issues from Ochanomizu University, we are conducting joint research to enable quantitative and objective gender equality measures using AI. We will transform ourselves toward better human capital management enabled by using the research outputs.
Serial articles: Required for those who are involved in AI! Think about AI ethics
We would like everyone to know about "what kind of AI ethics should we think about" in an easy-to-understand manner with examples.
