language:
- en
Annotated version of DEFCON 31 Generative AI Red Teaming dataset with additional labels for attack types.
This dataset is an extended version of the DEFCON31 Generative AI Red Teaming dataset, released by Humane Intelligence.
Our team conducted additional labeling on the accepted attack samples to annotate:
- Attack Targets (e.g., gender, race, age, political orientation) → tta01/redteaming-attack-target
- Attack Types (e.g., question, request, build-up, scenario assumption, misinformation injection)
The purpose of this extended annotation is to better understand:
- Which types of individuals or groups are most vulnerable to LLM attacks
- What kinds of prompting strategies are most effective in eliciting harmful outputs
⚠️ This dataset is shared for non-commercial, academic research purposes only.
📊 Dataset Contents
- 2,673 attack samples (from accepted DEFCON31 entries)
- 2 sets of annotations per sample:
attack_target: 7 superclasses and 19 subclassesattack_type: 10 binary-labeled prompting strategies
📄 Related Report
This dataset was analyzed in the following technical report (Korean), to be published on TTA
📄 License
This dataset is distributed under the Mozilla Public License v2.0, in accordance with the original dataset license. All annotations are provided under the same terms.
🏢 Research Institution and Contributors
This dataset was developed by the Center for Trustworthy AI at the Telecommunications Technology Association (TTA), South Korea.
Lead Researcher
- Dr. Yeajin Shin (Center for Trustworthy AI, TTA)
Collaborating Researchers
- Prof. Kyungsik Han (Hanyang University)
- Taehyung Noh (Hanyang University)
- Mingon Jeong (Hanyang University)
🙏 Acknowledgements
This work was supported by the Ministry of Science and ICT (MSIT) of Korea, as part of the “Establishing the Foundation of AI Trustworthiness” project, conducted by TTA.
We gratefully acknowledge the original organizers of the Generative AI Red Teaming Challenge:
- Dr. Rumman Chowdhury and Jutta Williams (Humane Intelligence)
- Sven Cattell (AI Village)
- Austin Carson (Seed AI)