Prompt Injection V2 Categorized is a dataset by NeuralAlchemy providing a multi-label taxonomy for prompt injection attacks. It is described as the definitive successor to a popular earlier dataset and offers a 6-dimensional classification system. The dataset was last updated on June 13, 2026.
Use Cases
- Train multi-label classifiers for prompt injection detection based on the described 6-dimensional taxonomy.
- Benchmark model robustness against different categories of adversarial prompts.
- Analyze the distribution and characteristics of prompt injection techniques.
- Develop fine-grained safety filters for AI applications based on threat categories.
Strengths
- Provides a 6-dimensional taxonomy, moving beyond a simple binary safe/unsafe label.
- Is described as the definitive successor to a highly popular earlier dataset.
- Has been demerged into clean, individual subsets for focused analysis.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- NeuralAlchemy
- Freshness
- Last updated 2026-06-13 08:00:37; freshness should be verified.