Guardrails Hardening and AI Governance

In this module of the Advanced AI Security Part 1 learning path, you will practice guardrails hardening on MathGPT with MATHGPT_GUARDRAILS on. Keyword injection should block. Path traversal should refuse. Arithmetic should still work. You will write a model card and a residual-risk note. Quotas, HTML sinks, and indirect injection remain accepted leftovers until Advanced AI Security Part 2. Upon completion of the module, students will be able to: - Prove guardrails true on healthz. - Show blocked injection and refused MCP traversal. - Deliver a one-page model card a named owner could sign. For inquiries, please write us at https://www.darkrelay.com/cybersecurity-course-inquiry