Microsoft's Newest Open-Source Project: Runtime Security For AI Agents
…These risks include goal hijacking, tool musiuse, identity abuse, supply chain risks, code execution, memory poisoning, insecure communications, cascading failures, human-agent trust exploitation, and rogue agents. This new Agent Governance Toolkit…