Skip to content
aiqbench takeAI Safety

aiqbench analysis

This article details critical security flaws in Microsoft Copilot where attackers can exfiltrate data from connected accounts via single-click exploits or injected prompts that persist across sessions. The vulnerabilities highlight systemic risks in how AI assistants handle external inputs and manage memory state, potentially allowing covert surveillance of user activity without triggering standard alerts. While the immediate threat is patched, these findings underscore urgent needs for robust input validation and ephemeral memory management in generative AI products.
Sources coveringCybersecurity News, LessWrong
What this isA model's reading of the article and related coverage, written for aiqbench. Not the publisher's words.

Critical Microsoft Copilot CoSnitch Vulnerability Lets Attackers Steal Sensitive Data With One Click

Cybersecurity News3 min read

A critical vulnerability in Microsoft Copilot Personal, tracked as CVE-2026-24301 and nicknamed CoSnitch, lets attackers silently siphon sensitive data from a victim’s connected accounts with nothing more than a single click on a malicious link. The flaw discovered by Varonis Threat Labs has already been patched by Microsoft as of August 18, 2026, but the […] The post Critical Microsoft Copilot CoSnitch Vulnerability Lets Attackers Steal Sensitive Data With One Click appeared first on Cyber Security News.

A critical vulnerability in Microsoft Copilot Personal, tracked as CVE-2026-24301 and nicknamed CoSnitch, lets attackers silently siphon sensitive data from a victim’s connected accounts with nothing more than a single click on a malicious link. The flaw discovered by Varonis Threat Labs has already been patched by Microsoft as of August 18, 2026, but the way…

Excerpt shown under the publisher's RSS terms · 56 words · full text and figures are on the publisher's site

Related coverage