← all datasets

Camouflaged Jailbreak Prompts

Emerging
1papers using it
2025first seen

The 'Camouflaged Jailbreak Prompts' dataset contains 500 curated examples of prompts—400 harmful and 100 benign—used to evaluate the effectiveness of safety protocols in large language models against subtle adversarial attacks that embed malicious intent within seemingly benign language.

Papers using Camouflaged Jailbreak Prompts (1)

Camouflaged Jailbreak Prompts — datasets — cybersecurity