ShieldGemma 9B is a specialist, not a generalist — it's built specifically to evaluate content safety rather than generate creative or analytical responses. It operates as a content classifier, assessing whether inputs or outputs violate safety policies across categories like harassment, hate speech, and dangerous content. Its narrow focus means it excels at its designated role but isn't designed for general-purpose conversation or reasoning tasks.