Dataset for "Jailbreaking LLMs for Creating Offensive Security Artifacts"

The dataset contains the complete result from the TFM "Jailbreaking LLMs for Creating Offensive Security Artifacts" (Universidad Carlos III de Madrid, 2025). It includes results containing prompt id, model, model version, developer, prompt type, requested content type, prompt, the three responses from each run, the rating of each run, average score of the three runs, month of test, notes and if the response was blocked.Disclaimer: This dataset contains prompts and model outputs that simulate malicious content. These were generated and collected solely for academic research in the context of the Master’s thesis. They are provided for analysis and reproducibility purposes only, not for practical use.