Beyond vibe coding: Implementing RBAC and guardrails for autonomous AI agents

In 2026, the rise of AI agents has shifted the platform's role from enabling human developers to governing autonomous entities. Without strict guardrails, an agent with "god-mode" access can cause catastrophic architectural drift or massive cloud waste. In this talk, Abhinav breaks down the technical framework for "Agentic Golden Paths." Attendees will learn how to implement granular RBAC specifically for non-human identities and how to treat AI-generated infrastructure as high-risk input within a GitOps workflow.

Key takeaways include:
• Building "cost-aware" guardrails
• Using OPA (Open Policy Agent) to validate AI-generated IaC scripts before deployment
• A blueprint for a "human-in-the-loop" approval gate that doesn't kill agentic velocity.