Federated Agent Trust, Shadow AI Discovery, and the Governance Stack in 2026

Tools

AGT v3.1.0 Ships Shadow AI Discovery, Agent Lifecycle Management, and Quantum-Safe Signing

The v3.1.0 release adds an 8-state agent lifecycle model with automatic credential rotation, a shadow AI scanner that finds unregistered agents in GitHub repos and running processes, and ML-DSA-65 post-quantum signatures alongside Ed25519. The PromptDefenseEvaluator covering 12 injection vectors is a practical addition for teams that need to audit their prompt attack surface.

Microsoft Agent Framework 1.0: Stable APIs, Multi-Agent Workflows, and A2A Support

Agent Framework 1.0 locks stable APIs across .NET and Python, ships graph-based workflow orchestration with checkpointing, and adds first-party connectors for OpenAI, Anthropic, Gemini, Bedrock, and Ollama. The GitHub Copilot SDK and Claude Code SDK integrations let you drop a coding-capable agent into a multi-agent workflow without rebuilding the agent loop yourself.

Analysis

Microsoft Agent Governance Toolkit: The Full Architecture Story

The seven-package toolkit maps directly to all 10 OWASP Agentic AI Top 10 risks with sub-millisecond policy enforcement, covering everything from goal hijacking to rogue agent kill switches. The OS-kernel and service-mesh analogies are not just marketing, the stateless policy engine and DID-based trust scoring are practical patterns builders can adopt incrementally.