AI has stopped being a checkbox feature in cybersecurity platforms — it's becoming the engine behind detection, investigation, and response. At the same time, the very LLMs and AI agents organizations are racing to deploy have opened a brand-new attack surface, from prompt injection to data leakage to compromised AI agents acting with excessive privilege. This session explores both sides of that coin: how leading security vendors are embedding AI and LLMs into their platforms to cut analyst workload and speed up response, and how a fast-growing "security for AI" category, AI-SPM, LLM firewalls, runtime guardrails, is emerging to protect the models themselves. Attendees will leave with concrete, vendor-agnostic use cases from both categories and a simple framework for evaluating where their own organization stands.