AIsecurityDojo uses documented, repeatable, and risk-aware methods when evaluating security controls, products, configurations, and AI system behavior.
Testing scope
Testing may include configuration review, feature validation, documentation review, permission analysis, attack-surface analysis, security-control verification, safe adversarial prompts, sandbox behavior, authentication and authorization checks, logging review, data-flow analysis, and comparison against published security guidance.
Safe testing
We do not intentionally perform unauthorized testing against third-party systems. Examples and demonstrations are designed for defensive education, controlled environments, systems we are authorized to evaluate, or publicly documented behavior.
Product testing
When hands-on access is available, we may test installation, configuration, workflows, integrations, security controls, reporting, and usability. When hands-on access is not available, an article may rely on documented capabilities and will be written accordingly rather than implying tests that were not performed.
Version sensitivity
AI products and security tools change rapidly. Test results may depend on product versions, model versions, settings, policies, operating systems, deployment environments, or subscription plans. Readers should verify current behavior before relying on an older result.
Reproducibility
Where useful, we aim to describe enough context for readers to understand what was evaluated and reproduce relevant defensive checks in their own authorized environment.