On October 6, 2025, KAS, together with the TUM Think Tank’s Civic Machines Lab and TUM Asia, hosted the seminar “AI and Civic Machines: Governance, Security, and Society” in Singapore. The event gathered experts from academia, industry, and policy to discuss how AI agents are reshaping work, governance, and urban life, with a focus on people-centred and responsible innovation.
Following the welcome remarks by Fabian Wagener, Director of the Media Program Asia at KAS Singapore, Prof. Dr. Orestis Papakyriakopoulos, Professor of Societal Computing at the TUM School of Social Sciences and Technology, opened the seminar with the first presentation. He emphasized the potential of AI agents to serve as partners in the workplace and underlined the importance of tackling barriers such as ethical dilemmas, data liability, and institutional inertia, while developing sector-wide and practitioner-driven recommendations.
The second presentation titled “Harnessing Agentic AI for Urban Transformation: From Smart Cities to Cognitive Cities” was delivered by Ajaz Munsiff, Chief Digital Officer for AI and Digital Transformation at Dell Technologies Asia-Pacific. He outlined how agentic AI can drive the transition from reactive, siloed smart cities to proactive, interconnected cognitive cities. Drawing on his industry expertise, he highlighted the role of AI-enabled systems in strengthening governance, resource management, and citizen engagement. A key element is AI-driven transformation at the edge, where computing power is moved closer to sensors and local infrastructure, enabling real-time decision-making, greater efficiency, and enhanced data privacy. This, he argued, lays the foundation for civic machines that adapt dynamically to societal needs.
The third presentation was delivered by Prof. Mohan Kankanhalli, Provost’s Chair Professor at the National University of Singapore (NUS) and Director of the NUS Artificial Intelligence Institute. In his presentation, “Using Persona to Bully an AI Model,” he introduced persona-driven probes as a way to test the vulnerabilities of large language models (LLMs). By simulating different human personality traits, the approach demonstrated how AI systems may respond under adversarial pressure, revealing that certain personas can make models more prone to unsafe and biased outputs. His thought-provoking talk underscored both the risks of manipulative interactions with AI and the importance of rigorous testing to ensure that such systems remain safe, accountable, and resilient in real-world applications.
Following the individual presentations, Anish Mishra, Project Manager and Research Fellow at KAS Singapore, moderated a panel discussion with the three speakers.
The central theme was the distinction between generative AI and agentic AI: The panellists explained that generative AI focuses on producing content or outputs within predefined tasks, while agentic AI is capable of pursuing complex goals in dynamic environments with limited supervision, thereby acting more autonomously and adaptively. The panel highlighted how this evolution raises pressing questions of trust, governance, and accountability, and stressed the need for international collaboration across academia, industry, and civil society.
The subsequent audience Q&A, followed by closing remarks from Dr. Markus Wächter, Vice President of TUM Asia and Executive Director of TUM Asia, and a networking reception, provided additional opportunities for exchange. These lively interactions underscored the importance of continued cooperation in steering AI development toward a sustainable and equitable future.
