TLDR Tech / InfoSec
Microsoft Open-Sources RAMPART and Clarity to Secure AI Agents During Development
Microsoft has released two open-source tools for testing AI agent security during development: RAMPART (Risk Assessment and Measurement Platform for Agentic Red Teaming), a Pytest-native framework for writing adversarial and benign safety tests covering harm categories like cross-prompt injection and data exfiltration, and Clarity, a "structured sounding board" that pressure-tests design assumptions before code is written. RAMPART builds on Microsoft's PyRIT and connects to an agent through a single adapter, evaluating test outcomes and reporting results. Where PyRIT targets black-box discovery after a system is built, RAMPART runs during development and Clarity captures design intent, turning red-team findings into reusable engineering assets across the agent lifecycle.
Read the full story