Last week, Robust Intelligence hosted the inaugural AI Risk Management Summit in Tokyo. The conference brought together leaders and experts from various industries to discuss the risks and security challenges introduced by the rapid adoption of AI, and to determine how Japanese companies might proactively mitigate these issues.
Executives from 46 organizations across a number of industries participated in the Summit, including representatives from Deloitte, Mercari, Mitsubishi, Mizuho Financial Group, Rakuten, Recruit, Sompo, Softbank, and Yahoo! Japan.
The highlights below offer a glimpse into the future of AI adoption and risk management in Japan.
The Summit commenced with introductory remarks from Masaaki Taira, a member of the Japanese House of Representatives and Chairman of the Liberal Democratic Party’s Project Team on the Evolution and Implementation of AI.
In his message, Taira shared his enthusiasm and support for the burgeoning AI industry. He underscored the potential of AI to greatly aid in addressing various issues faced by the country today and continuously emphasized the importance of responsible AI risk management.
Keynote: “What AI risks do we face now?”
In his keynote address, Yaron Singer, founder and CEO of Robust Intelligence, introduced a framework for understanding AI risk and discussed how businesses are working to address these different forms of risk.
His presentation began with an exploration of the three primary AI risk categories—operational, ethical, and security—with supporting examples provided for each. Yaron continued with four practical steps that Robust Intelligence recommends businesses consider to minimize risk in their deployment and continued use of AI:
- Policy: Identify AI vulnerabilities, consider how to control the identified risks, and formalize an internal policy.
- Testing: Implement testing tools to automate AI risk verification throughout the development and operation of AI models and services.
- Protection: Introduce an AI Firewall to prevent problematic inputs and outputs in real time and protect AI models and services in production.
- Reporting: Automate the reporting of detected risks to reduce costs and maintain communication between relevant departments.
Panel discussion: “Ensuring trust in AI: turbulent policy trends and the governance models required of companies”
In this panel discussion, four experts on AI risk management convened to share their perspectives on domestic and international policy and conceptualize a shared vision for AI governance standards across society.
Speakers also went beyond policy to examine risk within the AI supply chain. They noted how every player must consider the issue of AI governance, from model developers and AI service providers to the end users, in order to understand and combat AI risk.
Establishing the AI Governance Association
An especially important and exciting part of this panel was the announcement of the AI Governance Association (AIGA) establishment. The AIGA isa consortium where representatives across industries can discuss the state of AI governance so that Japanese companies and society can utilize AI with confidence and achieve sustainable growth.
Robust Intelligence will serve as the administration of AIGA and contribute to activities such as fostering a common understanding of AI governance in Japan and making proposals for policy makers. Over twenty-five companies have joined to date, including Sompo, NEC, Manulife, Deloitte, Recruit, Fujitsu, MUFG, and Mizhuo.
Concluding thoughts & the future of AI in Japan
Kojin Oshiba, the co-founder of Robust Intelligence, delivered closing remarks for the conference and reflected on the day’s discussions around AI risk management, policy development, and AI governance.
The emergence and proliferation of AI over the past year has undoubtedly taken the world by storm; the convening of Japanese business leaders and experts at the AI Risk Management Summit only underscores a shared desire to embrace AI as safely and securely as possible. Moreover, it highlights the need for collaboration in this space as no single entity can effectively contend with the challenges of this field alone.