A critical bug in Microsoft 365 Copilot, tracked as CW1226324, allowed the AI assistant to access and summarize confidential emails in Outlook’s Sent Items and Drafts folders, bypassing sensitivity labels and Data Loss Prevention (DLP) policies. Microsoft first detected the issue on January 21, 2026, with exposure lasting from late January until early to mid-February 2026. This flaw affected enterprise users worldwide, including organizations like the UK’s NHS, despite protections meant to block AI from processing sensitive data.
The vulnerability stemmed from a code error that ignored confidentiality labels on user-authored emails stored in desktop Outlook.When users queried Copilot Chat, it retrieved and summarized content from these folders, potentially including business contracts, legal documents, police investigations, and health records. Importantly, the bug did not grant unauthorized access; summaries only appeared to users already permitted to view the mailbox. However, feeding such data into a large language model raised fears of unintended processing or training data incorporation.
Microsoft swiftly responded by deploying a global configuration update in early February 2026, restoring proper exclusion of protected content from Copilot. The company continues monitoring rollout and contacting affected customers for verification, though no full remediation timeline or user impact numbers have been disclosed.As of late February, the patch was in place for most enterprise accounts, tagged as a limited-scope advisory.
This article has been indexed from CySecurity News – Latest Information Security and Hacking Incidents
Read the original article: