300 Agents, 60 People: How We Secured Our Own AI Agents - Aiden Sochowski, Runlayer
AI Engineer — Oct 27, 2026 — talk — 11.5K views · 48 likes
Why watch
A founding engineer at Runlayer shares real-world security lessons from operating 300+ AI agents across 60 people. The talk covers prompt injection vulnerabilities, agents being manipulated into approvals, cross-contamination between agents, and cost explosion incidents. Essential viewing for teams building or securing AI agent infrastructure in production.