The AI Agents You Didn’t Know You Had - Seguridad Mania.com - España y América Latina

AI agents aren't arriving through procurement. They're being built and deployed by your own marketing, engineering, and finance teams, acting on behalf of the humans who launched them. By the time security learns they exist, they've already been live for weeks. Security programs were built for human actors and deterministic systems. AI agents are neither. They make probabilistic decisions and can take actions at runtime that no pre-deployment review could have anticipated. Configuration audits and deployment checklists only capture an agent's posture at a single moment, but they can’t tell you everything it’s doing once it's running. Inventory starts to give you this view, but it’s got a blind spot. True AI governance requires going beyond inventory to runtime guardrails. You need more than inventorying agents or hardening pre-deployment policies, because the nature of agentic risk is in the execution, not the configuration. Security teams already monitor user behavior and enforce guardrails in real time; runtime security just extends that same model to agents operating at machine speed. In this session, Obsidian's Viet Tran and Tim Gohn walk through what that runtime security for AI agents looks like in practice and how security leaders can build towards runtime governance to enable AI adoption at scale. We'll cover: •The inventory gap: Why most programs stop at AI agent inventory and what that costs at runtime. •Runtime visibility: What it requires, and why, without it, governance is always a post-mortem. •A phased approach: How to move from observation to enforcement in a way that builds trust with the business, not friction. Whether your program is already managing agents in production or racing to get ahead of the ones appearing without review, this session defines what runtime governance looks like and what it takes to get there before something goes wrong.