AI agents automatically execute git malware when starting

AI agents execute malicious code in manipulated git repositories on startup – without user intervention, but with full rights ...
Infosecurity Magazine

NASA Ground Control Software Flaw Enables Unauthenticated Commands

A critical vulnerability in NASA's open-source AMMOS Instrument Toolkit (AIT)-GUI ground software has been found that could allow unauthenticated attackers to issue spacecraft and instrument commands, ...
The Hacker News

Marimo Notebook Flaw Could Run MCP Commands Before Cells Execute in Edit Mode

Marimo fixes CVE-2026-75149, an 8.7-severity flaw that can launch an MCP command before notebook cells run in edit mode.
Bleeping Computer

Critical Cisco UCCX flaw lets attackers run commands as root

Cisco has released security updates to patch a critical vulnerability in the Unified Contact Center Express (UCCX) software, which could enable attackers to execute commands with root privileges. The ...
Infosecurity-magazine.com

Cursor Security Bug Allowed Repositories to Execute Commands Before Trust Verification

A flaw in Cursor's command-line coding agent has been found to allow a cloned repository to run any command it chose on a developer's machine before they were asked whether they trusted it, and ...
The Hacker News

E4del and PINHOLE RATs Turn FTP Banners Into Dead Drops for Malware Commands

SOCRadar details E4del and PINHOLE RAT campaigns using FTP banners as dead drop resolvers to fetch commands and C2 details.