Microsoft 365 Copilot: Use Microsoft Agent 365 agents as connected agents in declarative agents

🚨 The Signal: Microsoft 365 Copilot now supports connecting declarative agents to Agent 365 agents via the Agent2Agent protocol. This expands AI agent interoperability, enabling more complex workflows and centralized management, but introduces new attack surfaces for agent-to-agent communication.

The Impact

Microsoft 365 administrators and developers are affected, facing increased risk from unmanaged agent interactions and potential data exfiltration.

• Administrators: New agent types require careful installation and permission management to prevent unauthorized access.
• Developers: Building declarative agents with A2A introduces new vectors for prompt injection and data leakage if not secured.
• Users: Access to unapproved or misconfigured Agent 365 agents could expose sensitive information or enable malicious actions.
• Security Teams: Need to monitor and audit inter-agent communications for anomalous behavior and policy violations.

The Action

1. Review and update your organization's AI governance policy to include Agent 365 and Agent2Agent protocol usage.
2. Implement strict access controls for installing and managing Agent 365 agents in the Microsoft 365 admin center.
3. Educate developers on secure coding practices for declarative agents, focusing on input validation and least privilege for agent interactions.
4. Establish monitoring and auditing for Agent 365 agent activities and inter-agent communications within Microsoft 365 Copilot.
5. Regularly review agent permissions and connections to ensure they align with business needs and security policies.

Domain: Agentic-AI · Impact: high · Workload: Other · Essential Eight: Restrict Administrative Privileges · ISM: ISM-0445, ISM-1175, ISM-1380, ISM-1507, ISM-1508, ISM-1509, ISM-1647, ISM-1648, ISM-1650, ISM-1686, ISM-1688, ISM-1689, ISM-1883, ISM-1897, ISM-1898