Ethical documentation of open-source AI models: a multivocal literature review and large-scale analysis of model cards [Dataset]

The dataset was collected using the official Hugging Face Hub API. It includes the 100,000 most-downloaded models available on the platform as of September 2025. For each model, structured metadata was retrieved and stored in tabular format, and, when available, the corresponding Model Card content was extracted and stored separately in JSON format. The dataset therefore includes: metadata_checkpoint.csv: File containing metadata for the 100,000 most-downloaded models from the Hugging Face Hub. This file comprises model identifiers, authors, creation and modification dates, number of downloads and likes, associated library, pipeline type, and descriptive tags. model_cards_checkpoint.json: File containing the textual content of 63,065 Model Cards stored in a single JSON file.