Secure AI agents in production | AgentOps Unlocked
Can you trust autonomous AI agents with real credentials? Learn how open source AgentOps infrastructure secures AI agents in enterprise production. Trusting the prompt is not a security posture. When AI agents execute tool calls, query databases, and interact with internal APIs over long time horizons, prompt engineering alone cannot prevent failures. In this kickoff episode of AgentOps Unlocked, Legare Kerrison breaks down how the agentic AI stack evolved and why traditional DevOps falls short for autonomous agents. We examine 4 critical infrastructure gaps: execution containment, identity, observability, and lifecycle governance. Learn how open source tools like Kubernetes, SPIFFE/SPIRE, and OpenTelemetry bring cloud-native operational rigor underneath autonomous agents. 00:00 Critical question in AI security 00:25 The evolution of the agentic AI stack 01:58 Bridging 4 critical AI infrastructure gaps 02:05 Execution containment and prompt injection 02:23 Identity, observability, and lifecycle governance 03:02 TLDR: Why prompt engineering fails at security 03:14 Open source tools for enterprise AgentOps Watch episode 2 of AgentOps Unlocked now → https://youtu.be/i4A8jN41NNs More AgentOps and AI resources: 📂 Access the AgentOps GitHub repository → https://red.ht/github-agent-ops 🤖 Learn about Red Hat OpenShift AI → https://www.redhat.com/en/technologies/cloud-computing/openshift/openshift-ai?sc_cid=RHCTG0260000496899 🔒 Explore Red Hat security solutions → https://www.redhat.com/en/topics/security?sc_cid=RHCTG0260000496899 #AgentOps #AIAgents #OpenSource #Kubernetes #OpenTelemetry #RedHat #EnterpriseAI