Data governance is usually conceptualized in terms of “privacy” v. “publicity”. Yet a core feature of pluralistic societies is association, groups that share with each other, privately. These are a diversity of (plural) publics, each externally private but with the ability to coordinate and share internally. Empowering plural publics requires tools that allow the establishment of shared communicative contexts and their defense against external sharing outside of context. The ease of spreading information online has challenged such “contextual integrity” and the rise of generative foundation models like GPT-4 may radically exacerbate this challenge. In the face of this challenge, we highlight why we believe the problem of “plural publics” to be a core challenge of data governance, discuss existing tools that can help achieve it and a research agenda to further develop and integrate these tools with a design eye specifically on the requirements of plural publics. READ THE PAPER.