Microsoft bug allowed Copilot to summarize confidential emails

2026-02-19 · Source: Dataconomy · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Fundamental Awareness, quick

Summary

Microsoft confirmed a bug, tracked as CW1226324, allowed its Copilot AI to summarize confidential customer emails without explicit permission, bypassing established data loss prevention policies. Active since January, the vulnerability specifically impacted Copilot Chat's functionality within Microsoft 365, enabling it to read and outline draft and sent emails marked with confidential labels. Microsoft initiated a fix rollout in February, following the initial report by Bleeping Computer. Copilot Chat is an AI-powered feature available to paying Microsoft 365 customers, integrating across Office applications like Word, Excel, and PowerPoint. Microsoft acknowledged that confidential email messages were "incorrectly processed" but did not disclose the number of affected customers.

Key takeaway

For security architects and compliance officers overseeing Microsoft 365 deployments, this incident underscores the critical need to audit AI service interactions with sensitive data. You should proactively verify that Copilot's data access adheres strictly to your organization's data loss prevention policies, especially concerning confidential information, and monitor for any unexpected data processing behaviors.

Key insights

A Microsoft Copilot bug exposed confidential emails by bypassing data loss prevention policies.

Principles

• AI systems can bypass security controls.
• Data loss prevention requires continuous validation.

In practice

• Verify AI integration with DLP policies.
• Monitor for unexpected data access patterns.

Topics

• Microsoft Copilot
• Data Loss Prevention
• AI Security Vulnerabilities
• Confidential Information
• Microsoft 365

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Security Engineer, Security Engineer, IT Professional

Related on AIssential

• Copilot searched your mailbox. LiteLLM handed out admin keys. Run this 5-check audit before your stack is next — Enterprise AI systems commonly lack trust boundaries for external input, creating systemic vulnerabilities across the stack.
• SearchLeak: How We Turned M365 Copilot Into a One-Click Data Exfiltration Weapon — AI-specific vulnerabilities can chain with classic web bugs to create novel, potent data exfiltration paths.
• Microsoft Copilot ignored sensitivity labels twice in eight months — and no DLP stack caught either one — AI assistants can violate trust boundaries invisibly within vendor inference pipelines, bypassing traditional security controls.
• Microsoft Hacked to Deliver Malware to Claude and Gemini Users — Microsoft's GitHub repositories were compromised to deliver credential-stealing malware to AI coding agent users.
• Microsoft Says Copilot AI Helped Knock Down Cybercrime Tools — Microsoft's Copilot AI reportedly aided in disrupting cybercrime operations.
• Critical Copilot vulnerability allowed hackers to steal 2FA code from users — AI models struggle to distinguish user instructions from malicious embedded commands, creating fundamental security vulnerabilities.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Dataconomy.