MultiPriv: A Multilingual & Multimodal Dataset for LLM Privacy Risk Research

MultiPriv is a dataset of Personally Identifiable Information entities and prompts designed for privacy risk research in large language models. It was created by author CyberChangAn and last updated on December 1, 2025. The dataset is multilingual and multimodal, though attribute-level VLM images are not directly included in the repository due to open-source certificate limitations.

Use Cases

• Benchmarking LLM privacy reasoning based on the described PII entities and prompts
• Evaluating adversarial prompt safety in multilingual contexts based on the dataset's scope
• Studying multimodal privacy risks for vision-language models based on the described image links

Strengths

• Dataset is multilingual and multimodal, covering multiple data types
• Focuses on PII entities, a specific and relevant domain for AI safety research
• Last updated date of 2025-12-01 suggests recent maintenance

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download
• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

CyberChangAn

Freshness

Last updated 2025-12-01 13:31:09; freshness should be verified