This dataset provides a cleaned and deduplicated collection of English-language social media posts designed for research on social media threat detection and classification. It supports binary threat detection and multiclass analysis of cybersecurity-related content across multiple social media platforms. The public release includes the main dataset, a data dictionary, a README file, and checksum information for integrity verification. This dataset is associated with the publication:Yasir A. Matnee and Saad Albawi, “A realistic multi-platform social media threat dataset,” IET Conference Proceedings, 2025. DOI: 10.1049/icp.2025.4375. Researchers using this dataset are requested to cite both the Zenodo dataset record and the associated publication.