Redefining Access Control: Scaling Policy as Code for Humans and AI Agents
Talkat KubeCon + CloudNativeCon Europe 2025 · on Thursday, 2025-04-03· at 16:45· Level 1 | Hall Entrance S10 | Room C·
summary
Argues that Policy as Code scales authorization to both humans and AI agents by enforcing a four‑perimeter pipeline: prompt filtering, RAG data protection, secure external access (MCP/gateways) and response enforcement. Demonstrates a Kubernetes agent built with PentiC AI (Cloud37 LLM) and Permit for object filtering, using Open Policy Agent/OpenFGA–style engines to evaluate user+action+resource policies and enforce namespace-level access and dry‑run warnings. Provides a repo/QR and claims this approach supports enterprise-scale, even “billions” of secured interactions.
AI-written summary · about
chapters (12)
- 0:00Welcome
- 1:48Access control for AI
- 7:11Access control layers
- 8:42AI agents and identity
- 11:13Four-perimeter framework
- 12:37Prompt filtering with agents
- 15:07RAG data protection
- 17:18External service access
- 19:03Policy on AI responses
- 20:50Kubernetes access control
- 28:15Future of AI access
- 29:39Key takeaways and closing
related talks
20mFrom Alignment to Access Control: A Unified View of GenAI...
27mSecurity-by-construction: How to weave authorization into modern app stacks using Open Policy Agent
43mApplication Code of Conduct - Full-Stack Policy as Code
15mWhy Security Matters: The Risks of Agentic AI and How to Mitigate Them
40mIn AI We Trust? Securing the Future, One Agent at a Time
35mImplementing AI RMF with Policy-as-Code Automation
31mAccess Policies for Hybrid Cloud Environments
38mPolicy as Code with Open Policy Agent