At ACM FAccT, we demonstrated why AI guardrails need the same scrutiny as models. Moving from static policies to context- and language-specific evaluations, our hands-on session proved that agentic guardrails equipped with tools like web search are vital for reliable, real-world deployment.