Defending Agentic AI

AI agents are no longer just assisting SOC analysts — they’re starting to take action. In this video, I break down what it actually means to defend AI agents as they move from investigation into containment and enforcement inside real security operations. This isn’t about building smarter models. It’s about securing autonomy — preventing over-trust, stopping prompt and log injection, and ensuring AI-driven decisions remain evidence-based and under human control. If you want to stay relevant as AI becomes embedded in the SOC, you need to understand how these agents can fail — and how to secure them. Don’t just use AI in security. ─ *Master it. Defend it* Chapters: 00:00 Introduction 01:30 The Current Debate 04:34 Agentic AI Vulnerabilities 07:31 Agent Autonomy 09:15 Demo Agent Vulnerabilities 13:01 Isolation Trust Model 16:34 Prompt Injection 19:35 Isolation Decision Workflow 23:37 Prompt Injection Remediation 24:46 Isolation Decision Remediation 25:53 New Cyber Security Roles 27:04 Closing Insights 📂 Explore my Cybersecurity Portfolio: https://lnkd.in/e7CfGQvj 💬 Connect with me on LinkedIn: https://www.linkedin.com/in/raymond-gonsalves/ 💻 Source Code: https://github.com/raymondgonsalves 🔹MAESTRO Framework Risk Mapping 🔹Security Audit Findings 🔹ORIGINAL vs REMEDIATION ISOLATION DECISION WORKFLOW 🔹Presentation Slides The above documents can be found at: https://github.com/raymondgonsalves/Defending_Agentic_AI/tree/main/docs 🎥 Mastering SOC Agentic AI: https://www.youtube.com/watch?v=Yxt6IEbwOWs #CyberSecurity #AICyberSecurity #SOCAnalyst #ThreatHunting #AgenticAI #SecurityAutomation