Turn specs into evals for any agent with ASSERT

This collection of Microsoft Security Blog posts covers emergent AI and supply-chain threats and defensive guidance. Key findings: ASSERT — an open-source framework to convert natural-language behavioral specs into executable evaluations for models and agents; a telemetry-driven playbook for reconstructing AI activity in Microsoft 365 Copilot and Azure AI to speed investigations and assess data exposure; threat actors exploiting AI branding as social-engineering lures; a prompt-injection vector in the Claude Code GitHub Action that could expose workflow secrets (responsible disclosure and Anth