license: mit
task_categories:

• text-classification
• other
  tags:
• cybersecurity
• risk-management
• GRC
• ai-safety
• red-teaming
  language:
• en
  size_categories:
• n<1K

AI Threat Vectors & Risk Register

Dataset Summary

This dataset is a comprehensive collection of AI-Related Threat Vectors, curated for GRC (Governance, Risk, and Compliance) leaders, AI Security Architects, and Red Teamers. It maps potential threats across the AI lifecycle, categorizing them by intent, agent, and system layer.

It is designed to assist in:

• AI Risk Assessments (NIST AI RMF, EU AI Act compliance).
• Threat Modeling (STRIDE for AI).
• Security Awareness Training.

Dataset Structure

The dataset contains the following fields:

• Intent: Whether the threat is Adversarial / Intentional or Negligence / Unintentional.
• Agent: The source of the threat (Internal employee vs External attacker).
• Category: The domain affected (Data, Model, System, Human).
• Threat vector: The specific name of the attack or failure mode (e.g., Data Poisoning, Prompt Injection).
• Explanation: A detailed description of how the threat manifests.
• Mitigation: (New) Strategic controls and technical measures to reduce risk.

Use Cases

For GRC Leaders

Use this dataset to populate your internal Risk Register. You can map these threats to your organization's controls (ISO 42001 or NIST AI RMF).

For Red Teamers

Use the Explanation and Threat vector columns to generate attack scenarios for table-top exercises or penetration testing scopes.

Example Data

Maintenance

This dataset is maintained by hewyler and is updated to reflect the evolving AI threat landscape (OWASP Top 10 for LLM, MITRE ATLAS).