Microsoft open-sources RAMPART and Clarity: Standardizing security processes for AI agents
- 14 hours ago
- 2 min read
Microsoft officially open-sourced two new security tools named RAMPART and Clarity. This is the tech giant's effort to help developers control risks from AI agents during the software development process, rather than only testing at the final step before deployment.

Clarity – A "thinking partner”
Unlike conventional testing tools, Clarity focuses on the early stages of a project. Microsoft describes this tool as an "AI thinking partner," helping engineers and product managers verify design assumptions before writing the first line of code.
This tool guides teams through structured conversations to clarify problems, explore solutions, and analyze failures. Decisions made through these conversations are stored directly in the project folder, ensuring transparency in traceability and the rationale behind design changes. The goal of Clarity is to prevent costly mistakes at the conceptual stage, where changing course is still less expensive.
RAMPART – A continuous security testing “engineer”
While Clarity is for thinking, RAMPART (short for Risk Assessment and Measurement Platform for Agentic Red Teaming) is the execution tool. Built on the PyRIT platform, RAMPART allows engineers to integrate safety tests into the Continuous Integration (CI) pipeline.
The strength of RAMPART is its ability to transform complex security vulnerabilities into repeatable CI tests. The tool focuses specifically on prompt injection attacks—a significant risk when AI agents have access to corporate data or perform actions on behalf of users. By turning lessons from Red Teams into engineering assets, RAMPART helps ensure that once a security bug is fixed, it never recurs.

A strategy toward safety engineering discipline
Microsoft's release of these two tools marks a major shift in the approach to AI safety. Instead of viewing security as an administrative hurdle at the end of a project, Microsoft is making it an integral part of the engineering process.
In a context where current AI systems do not just answer questions but can also write code, access records, and perform transactions, a small error can lead to serious consequences. The combination of Clarity and RAMPART promises to create a new standard for building safer and more reliable AI applications.
Currently, both tools are available on open-source platforms for the global developer community to access and use.
References:










Comments