Microsoft has released the Agent Governance Toolkit, a new MIT-licensed open-source project aimed at runtime security for AI agents. It claims to be the first toolkit addressing all ten agentic AI risks identified by OWASP, including goal hijacking, tool misuse, identity abuse, and rogue agents. The toolkit includes components such as Agent OS (policy engine), Agent Mesh (agent-to-agent communication security), Agent Runtime (dynamic execution rings), Agent SRE, Agent Compliance, Agent Marketplace, and Agent Lightning. It supports Python, Rust, TypeScript, Go, and .NET.
Sort: