A recent security flaw in Microsoft 365 Copilot is raising concerns over email privacy. The AI assistant is reportedly bypassing confidentiality sensitivity labels, leading to unauthorized summarization of potentially sensitive emails. This issue poses a significant risk to data protection within organizations.
Details of the Identified Flaw
First reported on February 4, 2026, and tracked under reference CW1226324, the flaw allows Microsoft 365 Copilot’s ‘Work Tab’ Chat feature to summarize emails labeled as confidential. These actions occur despite the presence of Data Loss Prevention (DLP) policies designed to restrict such processing.
Microsoft’s investigation revealed the root cause to be a code-level defect. This defect mistakenly allows the AI to access emails stored in the Sent Items and Draft folders, effectively bypassing the confidentiality labels intended to protect these messages.
Impact on Regulated Industries
The flaw is particularly concerning for sectors such as healthcare, finance, and government, where stringent email confidentiality is not just a best practice but a regulatory requirement. The National Health Service (NHS) has internally flagged the issue as INC46740412, highlighting its potential impact on public sector users of Microsoft 365.
Microsoft has started deploying a fix as of February 11, 2026, targeting affected environments. However, the resolution process is ongoing, and the issue remains unresolved for some users. Organizations are advised to monitor updates and review Copilot activity logs for any unusual access to labeled content.
Ensuring Data Security and Compliance
The bypassing of DLP policies by an AI tool like Copilot highlights a critical security gap. These controls are essential for data governance, and their circumvention can undermine an organization’s information protection strategy. Until a full resolution is achieved, organizations handling highly sensitive communications might consider temporarily limiting Copilot’s access.
Microsoft anticipates releasing further updates by February 18, 2026, with the aim of providing a comprehensive remediation timeline as the situation evolves.
Stay informed by following us on Google News, LinkedIn, and X for the latest cybersecurity updates. Contact us for further insights or to share your stories.
