The dataset is a social network of over 17,000 individuals who lived during the so-called Neo-Assyrian period of Mesopotamian history, primarily in the eighth and seventh centuries BCE. The undirected network of individuals connected by co-occurrences in cuneiform documents was semi-automatically extracted from the 'Prosopography of the Neo-Assyrian Empire'. In addition to two weighted versions of the one-mode co-occurrence network, the dataset also contains a two-mode person-text network and rich metadata for each individual. For the first time, the dataset allows large-scale computational analysis of social structures in the Assyrian Empire. The data is primarily stored as plain text and CSV files, inviting scholars to further expand and enrich it. The scripts and files used for creating and standardizing the data are also available in the Zenodo repository.