Privasis-USA

Dataset Description: Privasis is a large-scale synthetic dataset consisting of diverse text records—such as medical and financial records, legal documents, emails, and messages—containing rich, privacy-sensitive information. Each record includes synthetic profile details, surrounding social context, and annotations of privacy-related content. All data are fully generated using GPT-OSS-120B, supplemented with the personas from the public Nemotron-Personas-USA dataset (https://huggingface.co/datasets/nvidia/Nemotron-Personas-USA) The dataset is designed to support the training and evaluation of models or agents that operate on privacy-sensitive data. For example, it includes annotated text-sanitization instructions along with their corresponding sanitized outputs. The current release focuses on English-language content. This dataset is ready for commercial/non-commercial use. Dataset Owner(s): NVIDIA Corporation Dataset Creation Date: March 5th, 2026 License/Terms of Use: This dataset is licensed under the NVIDIA Dataset License Agreement. Additional Details This dataset contains synthetic data generated using GPT-OSS-120B Intended Usage: Researchers who would like to conduct privacy-related or social data-related projects. Individuals who would like to sanitize their private information from texts. Dataset Characterization Data Collection Method [Synthetic] Labeling Method [Synthetic] Dataset Format Text Records Main Corpus Columns The dataset has a single `corpus` configuration with the following columns: | Column | Type | Description | |---|---|---| | `id` | `str` | SHA-256 hash identifier for the record. | | `record_tags` | `list[str]` | Category tags for the record. | | `record` | `str` | The generated text containing PII and sensitive attributes. | | `profile` | `str` | JSON string of the synthetic person profile. | | `background_context` | `str` | Narrative context for the record. | | `record_type` | `str` | Description of the document type. | | `record_format`