This article examines the security implications of the Model Context Protocol (MCP) sampling feature in the context of a widely used coding copilot application.
MCP is a standard for connecting large language model (LLM) applications to external data sources and tools. We show that, without proper safeguards, malicious MCP servers can exploit the sampling feature for a range of attacks. We demonstrate these risks in practice through three proof-of-concept (PoC) examples conducted within the coding copilot, and discuss strategies for effective prevention.
Read more…
Source: Palo Alto Unit 42
Sign up for the Cyber Security Review Newsletter
The latest cyber security news and insights delivered right to your inbox
Related:
- UK Visa Portal exposed thousands of applicants’ passports and selfies — then called the lawyers on us
May 27, 2026
A website called UK Visa Portal publicly exposed thousands of passports and selfie photos of applicants who paid the site to obtain a U.K. immigration visa. An anonymous person notified TechCrunch about the security lapse, saying that the website was exposing at least 100,000 documents from people who uploaded their passports and selfies to the website ...
- Ghost CMS flaw hijacked to target hundreds of websites with ClickFix attacks
May 26, 2026
A critical-severity vulnerability that reportedly was patched three months ago is being exploited in a massive ClickFix campaign, researchers have claimed. In mid-February 2026, a critical SQL injection vulnerability was found in Ghost CMS, a popular open-source Content Management System (CMS) currently used by more than 57,000 websites, including the likes of 404 Media, The Canadian ...
- Industrial robots targeted by malware, which could open them up to hacking
May 25, 2026
A critical command injection vulnerability has been discovered in Universal Robots PolyScope 5, the operating system whucg powers the company’s collaborative robots. The flaw, tracked as CVE-2026-8153, carries a CVSS score of 9.8 and affects all software versions prior to PolyScope 5.25.1. This vulnerability could lead to complete compromise of the robot controller, affecting the confidentiality, integrity, and availability ...
- Another major Linux security flaw revealed — nine-year old issue could spell disaster for users
May 23, 2026
Security researchers Qualys discovered a major flaw in the Linux operating system (OS) that could let any ordinary user, or malicious actor, gain full admin access on vulnerable endpoints. This bug lingered in Linux systems since 2016, and affects the default installations of several major distributions, including Red Hat, SUSE, Debian, Fedora, AlmaLinux, CloudLinux, and others. Read more… Source: TechRadar News Sign up ...
- Cloud Atlas activity in the second half of 2025 and early 2026: new tools and a new payload
May 22, 2026
In 2025, Kaspersky observed pervasive SSH tunnel activity, which has remained active into 2026, affecting many government organizations and commercial companies in Russia and Belarus. Behind some of this activity is Cloud Atlas, a group which was known to Kaspersky researchers since 2014. During the investigation, the researches identified new tools used by this group, as ...
- ROADtools and Nation-State Tactics in the Cloud
May 22, 2026
ROADtools is a publicly available toolkit for offensive and defensive security purposes that attackers have integrated into cloud attacks. The tool is designed to: Enumerate Entra ID Register devices in Entra ID Acquire, exchange and manipulate Microsoft Entra ID tokens ROADtools is an open-source framework written in Python and built for red-teaming and research. It primarily targets the identity and ...