Guideline
Adversarial Evaluation Suite Development Guideline
This guideline provides an organization-neutral method for developing adversarial evaluation suites for AI models and AI systems. It supports repeatable testing for misuse resistance, data leakage, prompt injection, unsafe tool use, robustness, and lifecycle assurance.
- Type
- Guideline
- Version
- 1.0.5
- Date
- 2026-10-24
- Owner
- AI Security Maintainer
- Approving authority
- Governance Library Maintainer
- Classification
- Public
- Confidentiality
- Public
- Category
- AI Governance
- Review frequency
- 6 to 12 months and upon material AI threat, model, data, supplier, or assurance change
- Licence
- CC BY-SA 4.0
Read the full text on GitHub (opens in a new tab)
Outline
- Purpose
- Scope
- Development method
- Minimum test case fields
- Evidence requirements
- Informative references
- Limitations
Related documents
- AI Model Risk Frameworkai
- AI Model Risk Standardai
- AI Model Risk Assessment Procedureai
- AI Security and Risk Standardai
- Model Card Templateai
- System Card Templateai
- AI Testing, Validation and Documentation Standardai
Next
All Documents