melanieyes/

adaption-agent-security-behavior-v1

This dataset is a remastered version prepared using Adaption's Adaptive Data platform.

This dataset contains pairs of prompts and labels classifying AI agent actions in IT and security contexts as either benign or suspicious. The samples cover scenarios such as log analysis, code deployment, access control changes, and data handling to identify potential security violations or policy bypasses. Each entry evaluates whether the agent's behavior… See the full description on the dataset page: https://huggingface.co/datasets/melanieyes/adaption-agent-security-behavior-v1.