Fix formatting and clarify safeguards monitoring section

Author: v-mangrandhi

learn-pr/wwl-azure/protect-govern-ai-ready-infrastructure-azure/includes/4-implement-responsible-safeguards-content-filter.md

@@ -21,7 +21,7 @@ Beyond baseline filters, responsible AI practices such as documentation and tran
 - Model cards set realistic expectations about performance and limitations.  
 - Transparency notes support human oversight and responsible deployment.
 
-With safeguards are configured, operations teams establish continuous monitoring through Azure Monitor to track filter effectiveness and identify emerging threats. Content safety dashboards display key metrics: filter activation rates showing how often harmful content is blocked, category breakdowns revealing whether most violations involve hate speech or jailbreak attempts, and false positive rates indicating whether filters are too strict for your use case. You configure alerts that fire when activation rates exceed baseline thresholds—for example, notifying the security team when jailbreak attempts increase by 50 percent in a single day, suggesting coordinated abuse or a newly discovered vulnerability. Alert responses follow predefined procedures: temporary filter tightening to block suspicious patterns, stakeholder notification for transparency, and root cause analysis to understand whether incidents represent isolated events or systemic issues requiring architecture changes.
+With safeguards configured, operations teams establish continuous monitoring through Azure Monitor to track filter effectiveness and identify emerging threats. Content safety dashboards display key metrics: filter activation rates showing how often harmful content is blocked, category breakdowns revealing whether most violations involve hate speech or jailbreak attempts, and false positive rates indicating whether filters are too strict for your use case. You configure alerts that fire when activation rates exceed baseline thresholds—for example, notifying the security team when jailbreak attempts increase by 50 percent in a single day, suggesting coordinated abuse or a newly discovered vulnerability. Alert responses follow predefined procedures: temporary filter tightening to block suspicious patterns, stakeholder notification for transparency, and root cause analysis to understand whether incidents represent isolated events or systemic issues requiring architecture changes.
 
 These layered safeguards—dual checkpoint filtering, severity-based and custom content controls, transparency documentation, bias evaluation, and continuous monitoring—work together to demonstrate responsible AI operations. When compliance auditors ask how you prevent harmful outputs, you reference filter configurations and activation logs. When customers question whether your AI perpetuates bias, you present bias evaluation reports and mitigation strategies. When leadership asks whether AI risks could affect brand reputation, you show monitoring dashboards and incident response procedures. This evidence-based approach transforms responsible AI from an abstract principle into measurable operational practices.