Artificial intelligence (AI) offers transformative opportunities for cities, enabling data-driven decision-making, system optimization, and improved public service delivery. At the same time, AI represents disruptive change, introducing uncertainties and vulnerabilities that can exacerbate bias, inequality, privacy violations, and the concentration of power. Navigating these challenges requires focusing on ethical, social, and legal boundaries, as well as understanding the potential and limitations of current and future governance models and mechanisms. Anticipatory governance focuses on the capacity of institutions to foresee and prepare for emerging disruptions using for instance strategic foresight to provide a framework to address uncertainties and complexities. Foresight methods facilitate collective intelligence gathering, exploration of alternative futures, and long-term vision building. By actively engaging communities, civil society, and the private sector, these approaches support the co-creation of urban futures, help identify ethical redlines, and guide AI deployment in ways that are resilient, inclusive, and aligned with public values, while accounting for uncertainties, vulnerabilities, and the limits of governance. This study addresses a key knowledge gap in linking participatory scenario development to ethical considerations in urban AI. Based on focus group data, we develop two narrative scenarios that explore plausible to dystopian futures of AI in cities. By anchoring these scenarios in foresight methods, we illustrate how collaborative scenario development can identify ethical redlines and operationalize anticipatory governance, supporting resilient and inclusive urban futures and trajectories.