AI Sec Watch

Microsoft announced MDASH (a multi-model agentic scanning harness that uses over 100 specialized AI agents working together to find security vulnerabilities), which discovered 16 new vulnerabilities in Windows, including four critical remote code execution flaws (where attackers can run commands on systems they don't own). MDASH achieved an 88.45% score on a public cybersecurity benchmark, outperforming other systems, and is currently available only through a limited private preview program.

nnU-Net (a framework for automatically analyzing and segmenting images) had a vulnerability in its GitHub workflow where untrusted user input from issue titles and descriptions were sent directly to an AI agent without proper filtering. This allowed attackers to trick the AI agent into performing unintended actions like commenting on or relabeling issues, since the workflow ran automatically whenever someone opened an issue.

Meta is testing a feature on Threads that lets users tag a Meta AI account to answer questions or provide context in conversations, similar to how people use xAI's Grok on X. However, users discovered they cannot block the Meta AI account, which has caused frustration in the community.

Google is integrating Gemini, its AI model, deeply into Android and other devices as an 'intelligence system' that can automate tasks across multiple apps, understand what's on screen, and complete actions like booking reservations or building shopping lists. The move comes as Google competes with OpenAI and Anthropic for AI dominance, while also powering part of Apple's AI strategy, and represents a shift from traditional chatbots to agentic AI (systems that take actions on a user's behalf).

CVE-2026-42893 is a command injection vulnerability (a flaw where an attacker can insert malicious commands by exploiting how special characters are handled) in Microsoft 365 Copilot that allows an unauthorized attacker to tamper with data over a network. The vulnerability has a CVSS 4.0 severity rating (a moderate score on the 0-10 vulnerability severity scale). This issue was reported by Microsoft Corporation and published in May 2026.

Langflow (a tool for building AI-powered agents and workflows) has a path traversal vulnerability (a security flaw where attackers manipulate file paths to access files outside intended boundaries) in its Knowledge Bases API that allows authenticated attackers to delete arbitrary directories on the server by exploiting improper handling of knowledge base names. This flaw can cause data loss and service disruption.

CVE-2026-41614 is a vulnerability in Microsoft 365 Copilot for Desktop caused by improper access control (a weakness where the software fails to properly restrict who can do what), allowing an unauthorized attacker to perform spoofing (making something appear to come from someone else) on a local computer. The vulnerability has a CVSS 4.0 severity rating, though a full assessment from NIST has not yet been provided.

CVE-2026-41109 is a security flaw in GitHub Copilot and Visual Studio that allows an attacker to bypass a security feature by improperly handling special characters in output, which are then processed by another component (injection, where untrusted data is inserted into code or commands). The vulnerability can be exploited over a network by unauthorized attackers.

CVE-2026-41100 is a vulnerability in Microsoft 365 Copilot where improper access control (weak rules that don't properly check who should be allowed to do something) allows an authorized attacker to perform spoofing (impersonating someone or something else) on a local system. The vulnerability has a CVSS 4.0 severity rating (a moderate security concern on a 0-10 scale).