Microsoft Testing OpenClaw-Style Autonomous Agents Inside Microsoft 365 Copilot

Microsoft is integrating OpenClaw-style autonomous agent capabilities into Microsoft 365 Copilot, with corporate VP Omar Shahine confirming efforts to enable Copilot to “run autonomously around the clock” completing tasks on behalf of enterprise users. A separate enterprise-facing agent with improved security controls over the open-source OpenClaw is also in development.

OpenClaw’s open-source predecessor has well-documented security risks including prompt injection, uncontrolled tool invocation, and data exfiltration through agent-driven workflows. Security teams should review AI agent permissions and data access scopes before enabling autonomous Copilot features, establish approval gates for high-impact agent actions, and ensure audit logging is in place for automated task execution in Microsoft 365 environments.