AI agents now act with human credentials, permissions, and authority, at a speed no analyst can match. In 2026, attackers took notice, using a stolen OAuth token from a third-party AI app to reach Vercel’s internal systems and turning developers’ own coding agents into persistence for the keyv npm worm. Most teams are answering with another dashboard. In this session, the Exaforce product team will explain why AI security works best inside the SOC, where every agent action resolves to a real person, device, and resource, and then walk through real scenarios from a production environment showing how Exaforce finds, flags, detects, and contains AI agent threats with nothing new deployed on the endpoint. Whether you’re governing AI adoption across the business or triaging the alerts it creates, this session gives you a practical model for securing agents alongside everything else your SOC already protects.
Details
- Date: October 26, 2026, 11 am PT
- Length: 45 minutes
- Hosts: Kavita Varadarajan, Lahav Ezer
What you will learn
- Why AI and agent security belongs inside the SOC instead of in a separate tool
- How to inventory every agent, skill, and MCP server across your fleet without deploying anything new
- How Exaforce tells an agent’s action apart from a human’s, even when both use the same credentials
- What real agent threats look like in production, from leaked secrets in AI chats to a desktop agent stealing session cookies
- Where to automate response and where to keep a human in the loop