Microsoft has introduced the Agent Governance Toolkit, an open source project designed to monitor and control AI agents at runtime as enterprises seek to integrate them into productive workflows. The toolkit, which is a response to the Open Worldwide Application Security Project (OWASP)’s growing focus on security risks AI and LLM, adds a layer of runtime security that enforces policies to address issues such as prompt injection and improves visibility into agent behavior in complex, multi-step workflows.
More specifically, the toolkit maps the 10 risks OWASP, including goal hijacking, tool misuse, identity abuse, supply chain risks, code execution, memory poisoning, insecure communications, cascading failures, human-agent trust exploitation, and rogue agents.
See also: Contagious Interview – North Korea: 1,700 malicious npm, PyPI, Go, Rust packages
How Microsoft's Agent Governance Toolkit works
The rationale behind the toolkit, according to Microsoft executive Imran Siddique, comes from how AI systems increasingly resemble loosely controlled distributed environments, where multiple untrusted entities share resources, make decisions, and interact externally with minimal oversight.
This prompted Microsoft to apply proven design patterns from operating systems, service meshes, and reliability engineering to bring structure, isolation, and control to these environments, Siddique added. The result was the merging of these principles into a toolkit consisting of seven components available in Python, TypeScript, Rust, Go, and .NET.
The multi-language approach, according to Siddique, aims to meet developers where they are and enable integration into heterogeneous enterprise stacks. In terms of components, the toolkit includes modules such as a policy enforcement layer called Agent OS, a secure communication and identity framework called Agent Mesh, an execution control environment called Agent Runtime , and additional components, such as Agent SRE, Agent Compliance, and Agent Lightning, which cover trust, compliance, market governance, and reinforcement learning oversight.
See also: Zero-day attack targets Adobe Reader users
Beyond its modular design, Siddique further wrote that the toolkit is built to work with existing development ecosystems: “We designed the toolkit to be framework-agnostic from day one. Each integration connects to a framework’s native extension points, LangChain’s callback handlers, CrewAI’s task decorators, plugin system , Microsoft’s Agent Framework middleware pipeline, so adding governance doesn’t require rewriting the agent code.”
This approach would reduce the cost and risk of integration, allowing developers to introduce governance controls into production systems without disrupting existing workflows or incurring the cost and complexity of refactoring applications.
See also: Iranian Hackers Target US Critical Infrastructure with PLC Attacks
Siddique even went as far as to provide examples of several framework integrations that have already been deployed in production workloads, including LlamaIndex's TrustedAgentWorker.
Those wishing to explore the toolkit, which is currently in public preview, should know that it is available under an MIT license and structured as a monorepo with independently installable components.
Microsoft, in the future, plans to move the project to a foundation-led model and is already working with agentic AI community to support broader governance and management.
🔒 Protect your privacy with Proton VPN
Swiss VPN from the creators of Proton Mail — strict no-logs policy, strong encryption, and built-in NetShield that blocks ads, trackers, & malware.
- ✔ No-logs, based in Switzerland (except 14-Eyes)
- ✔ NetShield: blocks ads, trackers & malicious domains
- ✔ Covers all devices — free version available
The link is an affiliate link — SecNews may receive a commission at no additional cost to you. It does not affect the independence of our article writing.
