Generative AI and Safety in Dynamic Relationship Contexts

Generative AI systems are rapidly reshaping how people communicate, seek advice, and make sense of social situations. Increasingly embedded in everyday life, these systems influence not only individual users but also the relationships, communities, and social norms within which those users are situated. Conversational AI is now routinely used for emotional support, conflict navigation, social coordination, and meaning-making. As a result, generative AI has become an active participant in social processes that carry real consequences for trust, well-being, and collective safety.This growing social integration introduces safety challenges that extend beyond individual-level risk models. When AI systems are used within dynamic relationship contexts (i.e., romantic partnerships, friendships, families, classrooms, or peer groups), their responses can shape power dynamics, reinforce or destabilize norms, and influence how responsibility, blame, and care are distributed. Harms in these settings often emerge incrementally and relationally, through patterns such as normalization of unhealthy dynamics, amplification of social pressure, reinforcement of one-sided narratives, or subtle misinterpretation of intent across repeated interactions.Despite these realities, most AI safety evaluations and governance frameworks continue to assume isolated, static users, overlooking how risk accumulates over time and propagates through social ties. This gap limits our ability to understand the broader societal impacts of generative AI, including its role in shaping interpersonal conflict, emotional labor, social learning, and collective decision-making.This topic invites contributions that examine how generative AI systems interpret, mediate, and influence ongoing social relationships, and how these influences scale from individual interactions to community-level and societal outcomes. We particularly welcome work that connects interactional AI behavior to broader implications for social trust, inequality, accountability, and harm prevention, and that considers how design and governance choices can either mitigate or exacerbate these effects.Relevant contributions may:• analyze AI behavior in relationally dynamic and socially embedded settings• identify unintended or unsafe response patterns that affect relationships, groups, or communities over time• examine how AI systems shape social norms, expectations, and decision-making within relationships• propose evaluation frameworks that capture longitudinal, contextual, and relational risk• develop design principles or governance mechanisms that support de-escalation, neutrality, and relational awareness• consider cultural, social, and developmental factors that influence how relational AI use is interpreted and negotiated.By setting dynamic relationships as a unit of analysis, this topic aims to broaden generative AI safety and design research to better reflect how AI systems participate in and transform real-world social systems, with consequences that extend beyond individual users to society at large. Topic editors: Chen Ling, Yuerong Hu, Kaiwen Sun, Xinyue Shen Source: https://www.frontiersin.org/research-topics/79646/generative-ai-and-safety-in-dynamic-relationship-contexts