Artificial intelligence is rapidly transforming how innovation is organised within university-centred ecosystems. As AI-enabled practices diffuse across teaching, research, and collaborative activities, universities are required to redefine their governance arrangements in order to preserve transparency, accountability, and institutional legitimacy. These tensions are particularly visible in Academic Innovation Ecosystems (AIEs), where universities act as orchestrators coordinating interactions, infrastructures, and knowledge flows among diverse internal and external actors. While the literature increasingly acknowledges the relevance of responsible AI in higher education, we still lack a process-oriented understanding of how universities translate abstract principles into operational governance configurations embedded in ecosystem activities.
This study develops a conceptual model of AI governance in AIEs through qualitative document analysis grounded in abductive theorising. The empirical basis consists of official AI regulations and policy documents issued by Italian universities in response to emerging European regulatory frameworks. Out of 95 universities, 14 had adopted formal AI-related guidelines. These documents were analysed alongside institutional strategies and mission-related materials to capture how AI governance is articulated across education, research, and third-mission domains. The university is treated as the unit of analysis in its role as ecosystem orchestrator, with attention to how governance provisions shape the coordination of data and decision processes across missions.
Preliminary findings reveal a clear concentration of governance efforts around generative AI in teaching and research, whereas third-mission and administrative applications remain comparatively underdeveloped. In educational contexts, universities primarily focus on academic integrity and disclosure practices, whereas research-related provisions tend to emphasise authorship, intellectual property, and compliance with data protection norms. Across cases, regulatory references to European frameworks are present but uneven, and the operationalisation of monitoring mechanisms often remains generic rather than process-specific.
These patterns are theoretically significant because effective ecosystem orchestration depends on governance arrangements that are embedded in concrete practices rather than articulated solely as high-level principles. When AI governance remains reactive and unevenly specified across missions, the coordinating role of the university risks becoming fragmented. The proposed model advances research on innovation ecosystems by linking institutional conditions, infrastructural arrangements, and mission-specific governance choices, offering a foundation for more accountable and sustainable AI adoption within academic settings.
