Regulating Reality: Exploring Synthetic Media Through Multistakeholder AI Governance

Abstract: Artificial intelligence's integration into daily life has brought with it a reckoning on the role such technology plays in society and the varied stakeholders who should shape its governance. This is particularly relevant for the governance of AI-generated media, or synthetic media, an emergent visual technology that impacts how people interpret online content and perceive media as records of reality. Studying the stakeholders affecting synthetic media governance is vital to assessing safeguards that help audiences make sense of content in the AI age; yet there is little qualitative research about how key actors from civil society, industry, media, and policy collaborate to conceptualize, develop, and implement such practices. This paper addresses this gap by analyzing 23 in-depth, semi-structured interviews with stakeholders governing synthetic media from across sectors alongside two real-world cases of multistakeholder synthetic media governance. Inductive coding reveals key themes affecting synthetic media governance, including how temporal perspectives-spanning past, present, and future-mediate stakeholder decision-making and rulemaking on synthetic media. Analysis also reveals the critical role of trust, both among stakeholders and between audiences and interventions, as well as the limitations of technical transparency measures like AI labels for supporting effective synthetic media governance. These findings not only inform the evidence-based design of synthetic media policy that serves audiences encountering content, but they also contribute to the literature on multistakeholder AI governance overall through rare insight into real world examples of such processes.