Privacy 0 sources · Running 87 days · Last updated Aug 18, 2026

Microsoft Copilot Security Flaw Exposes User Data

Why It Matters

This incident highlights the critical vulnerabilities in AI systems that can lead to significant privacy breaches, affecting countless users who rely on Microsoft's services. The exposure of sensitive information, such as 2FA codes, poses serious risks, including identity theft and unauthorized access to personal accounts. As AI technologies become increasingly integrated into daily life, ensuring their security is paramount to maintaining user trust and safeguarding personal data.

Summary

A significant security vulnerability was identified in Microsoft's M365 Copilot AI platform, allowing hackers to extract sensitive user data, including two-factor authentication (2FA) codes. The flaw arose from the AI's failure to differentiate between legitimate user commands and malicious requests embedded within third-party content. Researchers from Varonis demonstrated this vulnerability by manipulating an undocumented parameter in the Copilot's URL, leading to unauthorized access to confidential information without user consent. Despite Microsoft's efforts to implement security measures, the incidents have raised alarms about the effectiveness of these guardrails and the potential for further exploitation. Currently, Microsoft is working to address these vulnerabilities and enhance security protocols to protect user data.