Microsoft unveils Microsoft Security Copilot agents and new protections for AI

Microsoft announces an expansion of Security Copilot with six first-party AI agents (e.g., phishing triage, Purview alert triage, Entra conditional access optimization, Intune vulnerability remediation, and a threat intelligence briefing agent) and five partner-built agents, alongside new Purview data security investigations, Defender AI security posture management across Azure/AWS/GCP and multiple model catalogs, new detections for OWASP AI risks (indirect prompt injection, sensitive data exposure, wallet abuse), Entra internet access controls to curb shadow AI, Purview DLP in Edge to block sensitive data entry into AI apps, and phishing protection in Microsoft Teams, with previews and GA beginning April–May 2025.