7,725 real-world and research incidents involving generative-AI and agentic-AI systems collected by emmanuelgjr. Each incident is tagged with categories from industry frameworks including OWASP Top 10 for LLM Applications and OWASP Agentic Top 10. The dataset was updated on June 3, 2026.
Use Cases
- Analyze trends in prompt injection and jailbreak incidents based on the OWASP LLM framework tags
- Study patterns of data exfiltration and model compromise based on the incident descriptions
- Benchmark AI risk management frameworks based on the mapping of incidents to multiple industry standards
- Train models to classify AI security incidents based on the framework tags
Strengths
- Contains 7,725 documented incidents, providing a substantial corpus for analysis
- Each incident is mapped to four industry frameworks, including OWASP Top 10 for LLM Applications
- Includes both real-world events and research incidents, covering a broad scope of issues
Limitations
- Column-level documentation is absent; field semantics must be inferred after download
- Row count is known, but the specific incident types and their distribution are not detailed in the metadata
- Freshness should be verified; last updated 2026-06-03 17:00:35
Provenance
- Source
- huggingface
- Collection Method
- Likely curated from public reports and research papers.
- Freshness
- Last updated 2026-06-03 17:00:35