Oriel — AI & technology worth noticing, every day.

Necent/llm-jailbreak-prompt-injection-dataset

Dataset · https://huggingface.co/datasets/Necent/llm-jailbreak-prompt-injection-dataset

LLM Jailbreak & Prompt-Injection Dataset A unified safety dataset combining 30+ public sources for training LLM guardrails, content moderation classifiers, and response-safety filters. Schema (orthogonal multi-label, WildGuard-style) Instead of a single binary is_dangerous, every example carries four orthogonal labels matching the structure used by AI2 WildGuard, IBM Granite Guardian, and Azure Prompt Shields: Column Type Description prompt str The… See the full description on the dataset page: https://huggingface.co/datasets/Necent/llm-jailbreak-prompt-injection-dataset.

First seen by Oriel on .

AI summary

Necent/llm-jailbreak-prompt-injection-dataset unifies 30+ public safety datasets with orthogonal labels for training prompt-injection guards, content moderation, and response-safety filters.

  • Each example includes prompt, optional response and model metadata, source, language, attack technique, and separate harmfulness and adversarial labels.
  • prompt_harmful and prompt_adversarial are always set, while response_harmful and response_refusal are nullable for approximately 75–80% of rows.
  • The orthogonal labels distinguish harmful content from adversarial behavior, including cases that are harmful, adversarial, or both.
  • Sources cover jailbreaks, prompt injections, harmful behaviors, toxicity, multilingual safety, agentic attacks, and synthetic obfuscation such as base64, hex, ROT13, and leetspeak.
  • The dataset is intended for defensive AI safety research; integration code is MIT-licensed, while underlying sources retain their original licenses.

#llm safety #prompt injection #jailbreaks #content moderation

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Likes

Total likes reported by the source.

  • Hugging Face Datasets: 81, latest observation ; +7 over observed period since

Downloads (30d)

Rolling 30-day downloads. A change in this value is not new downloads during the observed period.

  • Hugging Face Datasets: 2,126, latest observation ; +722 over observed period since

Aliases

SourceExternal IDURL
Hugging Face DatasetsNecent/llm-jailbreak-prompt-injection-datasethttps://huggingface.co/datasets/Necent/llm-jailbreak-prompt-injection-dataset Read discussion