Agents watching agents: Microsoft strengthens defenses against the new wave of internal risks

Microsoft is enhancing its cybersecurity measures to address risks posed by autonomous AI agents. These agents require privileged access to corporate networks, making them potential internal threats. The company introduced Guardian Agents within the Project Perception to monitor and audit AI activities. The Microsoft Digital Defense Report 2026 highlights that malicious connections accounted for 52% of observed attacks. The new system isolates AI operations and includes mechanisms to assess agent behavior. While designed for large corporations, advanced controls will also be available for SMEs. The implementation of the NIS2 directive emphasizes the need for rigorous cybersecurity audits. Additionally, Microsoft is strengthening its Sovereign Cloud capabilities to address technological dependency concerns in the EU.