Marius Sandbu • 3/24/2026

Generative AI Agent Governance

This article discusses the governance of Generative AI agents, categorizing them into endpoint-based agents (e.g., GitHub Copilot, Claude Code), PaaS agents (e.g., Copilot Studio, Tasked), and framework agents (e.g., Langgraph, Microsoft Agent Framework). It examines common building blocks like LLM access, identity, skills/MCP, and data interaction. The article raises key questions about centralized monitoring and security for these agents, highlighting the need for governance in enterprise environments where AI agents interact with company data and systems.

0 comments

#llm #generative ai #AI Agents