Microsoft Introduces Tool to Automate Risk Discovery and Addressing in AI Agents

Microsoft has introduced a new tool called run-assert-eval, designed to automate the process of discovering and addressing risks in AI agents. The tool is an extension of Microsoft's earlier work on ASSERT and Agent Control Specification, which aimed to provide a more rigorous and portable way to enforce policy at runtime.

The current practice of using these tools assumes that teams have already written down their requirements and that someone has the time and expertise to connect every step by hand. However, this approach can be limiting as it relies on the completeness of the written requirements and the ability to manually integrate different components.

run-assert-eval addresses these limitations by providing a single prompt in VS Code that discovers risks, measures how often the agent fails, generates runtime policy directly from those findings, and reruns the same evaluation to prove whether the fix worked.