Microsoft on Thursday released an open source standard called Agent Control Specification (ACS) that aims to let developers, compliance teams, and security staff define consistent, granular rules for AI agents. The specification checks policy files at multiple interception points while agents run, so teams can control allowed actions, blocked actions, human approvals, and evidence logging.
What ACS does
ACS lets teams write policy files that state what an agent may do, must not do, when a human must approve an action, and what evidence must be logged for review. Those policy files are evaluated at defined interception points as the agent executes tasks, including before input processing, before calling a tool, after a tool returns a result, and before the agent’s final response.
Policy actions and enforcement
A policy can allow an action, block it, redact sensitive information, or require human approval. Developers can embed classifiers for inputs and outputs, include LLM-based “judges” with prompts to evaluate policy decisions, and add logic to check tool calls, tool selection, input accuracy, output usage, and responses.
Integration and portability
Because policies are single files, they can be bundled with agents, allowing the same security policy to travel with an agent across frameworks and environments. Microsoft says this design addresses fragmented controls that teams currently build via system prompts, custom application checks, and classifiers.
Why Microsoft framed it
Microsoft introduced ACS as enterprises increasingly deploy agentic AI across applications, workflows, and products, and as developers improvise controls to prevent tool misuse and unintended cascading failures. ACS aims to create a common governance layer that teams can reuse and audit across contexts.
SDK and ecosystem support
ACS is shipping as an SDK with plug-ins for LangChain, the OpenAI Agents SDK, the Anthropic Agents SDK, AutoGen, CrewAI, Semantic Kernel, Microsoft.Extensions.AI, MCP tools, and others. Microsoft intends the SDK and plugins to make adoption easier across agent frameworks.
Featured image credits: Roboflow Universe – Chayada
For more stories like it, click the +Follow button at the top of this page to follow us.