Microsoft has introduced the Agent Control Specification (ACS), an open-source standard designed to give developers a more consistent and granular way to control AI agents. The specification allows developers, compliance, and security teams to define policies that dictate what agents can and cannot do, when human approval is required, and what evidence should be logged for review. These policies are checked at multiple points during an agent’s workflow to ensure compliance with set guardrails. The move comes as developers struggle to manage AI agent behavior across different environments and workflows, often using ad-hoc methods that lack consistency and auditability. Source: techcrunch
ACS enables developers to embed policies into agents, allowing them to be reused across various frameworks and systems. The specification supports a range of actions, including allowing or blocking specific tasks, redacting sensitive information, and requiring human approval for certain actions. Developers can also integrate classifiers to assess inputs and outputs, add large language models (LLMs) as policy judges, and implement logic to verify tool calls and responses. This approach aims to create a unified governance layer that simplifies the management of AI agent behavior. Source: techcrunch
Microsoft is releasing ACS as an SDK with plug-ins for multiple platforms, including LangChain, the OpenAI Agents SDK, the Anthropic Agents SDK, AutoGen, CrewAI, Semantic Kernel, Microsoft.Extensions.AI, MCP tools, and more. The initiative is part of Microsoft’s broader efforts to enhance AI governance and ensure that AI agents operate safely and predictably in enterprise environments. Source: techcrunch