Designing Defense for Offense: Governance Issues for Companies in the Generative AI Era – Mitsui Sumitomo Bank, Toyota, Marubeni and Others Discuss Cross-Industry

From left: Yaguchi Hiromu (Researcher, Max Planck Institute for Human Development / Research Fellow, Anthropic), Masuda Masashi (Partner, Mori Hamada Matsumoto Law Firm / Visiting Professor, Rikkyo University), Tsuda Takuya (CEO, MQue).

MQue, a deep-tech company specializing in simulation, modeling, and conversational AI, held a roundtable event 'Future HR × AI Roundtable (2nd)' on May 22, 2026, themed 'Designing Defense for Offense in the Generative AI Era.'

MQue is a deep-tech partner that combines multiple advanced technologies for practical implementation, with world-class research expertise. In conversational AI, it focuses on unstructured data analysis and human digital twins, supporting strategic decision-making in recruitment, organization, and talent.

As the AI battlefield shifts from efficiency to advanced judgment, creativity, and dialogue, this closed roundtable aimed to share challenges in security, bias, ethics, and governance. Participants included practitioners from HR, legal, management planning, DX, risk management, and security at major companies like Mitsui Sumitomo Bank, Toyota, and Marubeni. Key discussion points included:

- How much judgment to delegate to AI and where human responsibility lies
- What data can be used and to what extent
- How to design 'defense' not as a constraint but as a foundation for safe advanced AI use

Special lectures were given by Masuda Masashi (lawyer) on 'Organization and Responsibility in the Advanced AI Era – Frontier of Technology and Law' and Yaguchi Hiromu on 'Generative AI Transforming Management and HR.' Panel discussions and table talks followed.

Yaguchi highlighted structural changes: traditional people analytics failed due to structured data limitations; generative AI now enables active collection of unstructured data (intentions, context), elevating HR to a strategic tool. He introduced 'Human Digital Twin' for probabilistic decision-making, e.g., potential-based promotion, reorganization success prediction, M&A synergy calculation. He emphasized that input data quality – especially organizational culture and tacit knowledge – is key to value creation.

Masuda addressed legal risk management: beyond binary risk assessment, companies must understand risk location, magnitude, probability, and sanctions, and design risk-taking aligned with business objectives. As AI evolves from output tools to autonomous agents, responsibility remains with humans. He advocated 'Human-in-the-Loop' design and establishing cross-functional AI governance committees with independence.

Panel discussions covered responsibility allocation, organizational structure for AI promotion, business process redesign, HR AI use, unstructured data collection and privacy. Management involvement was deemed essential, along with continuous monitoring. The risk of not using AI was acknowledged. In HR, preserving company-specific 'culture' while automating was a key question. Japanese companies' cross-departmental talent development was seen as having potential in the generative AI era, with the suggestion to 'retain strengths while moving forward.'