Microsoft open-sources two agentic AI safety tools

Microsoft has open-sourced two AI safety tools aimed at helping developers build more secure AI agents. RAMPART is a pytest-based red-teaming framework that embeds automated attack simulations into CI/CD pipelines, testing for threats like prompt injection and verifying agent behavior across hundreds of runs. It also helps teams reproduce security findings and validate mitigations. The second tool, Clarity, is an AI agent that acts as a pre-development sounding board, prompting teams with architectural and safety questions before any code is written. Microsoft says both tools reflect a shift toward treating AI safety as an engineering discipline rather than a philosophical concept.