Microsoft Copilot Bug Exposes Confidential Emails
A vulnerability in Microsoft's Copilot AI has reportedly resulted in the unauthorized disclosure of confidential Office emails. Users found that the AI-powered email summarization feature was including information from emails outside the intended search context. The flaw highlights the security risks of integrating generative AI into enterprise workflows, particularly how an AI's context window can inadvertently access more data than intended.
- The vulnerability was a code issue specifically within the "work tab" of Microsoft 365 Copilot Chat, which incorrectly processed emails from users' "Sent Items" and "Drafts" folders. - This bug allowed the AI to bypass established security protocols, including Data Loss Prevention (DLP) policies and "confidential" sensitivity labels that are designed to restrict access to sensitive information. - Microsoft's internal tracking number for the incident is CW1226324, and the company first detected the issue on January 21, 2026. - A fix for the vulnerability began rolling out in early February 2026, though Microsoft has not provided a specific timeline for when all systems will be patched. - The company has not disclosed how many organizations or users were affected by the breach, stating that the scope of the impact may change as their investigation continues. - Coincidentally, just days before Microsoft confirmed the bug, the European Parliament's IT department blocked built-in AI features on official devices, citing concerns that such tools could upload confidential correspondence to external cloud servers.
