Ghostcommit is a proof of concept that shows how AI assistants used to review software code can be tricked by hidden instructions embedded in images.
The academic ASSET Research Group showed that an attacker can place instructions inside an image file, point to it in an AGENTS.md file, and get an AI coding agent to follow those instructions during a later task.
A pull request is basically a formal “please review and add my changes” request that a developer sends before changes are added to the main version of a software project. Human reviewers and, increasingly, AI coding tools may review the changes before they’re accepted.
Read more…
Source: MalwareBytes Labs
Sign up for the Cyber Security Review Newsletter
The latest cyber security news and insights delivered right to your inbox
Related:
- Microsoft shuts down illegal code-signing operation used by ransomware criminals to mask their malware
May 19, 2026
Microsoft seized websites and took down hundreds of virtual machines running a cybercrime service that allegedly sold code-signing certificates to ransomware gangs, thus making their malware look like legitimate software – and allowing criminals to infect thousands of machines in the US, including at least 12 owned and operated by the Windows giant. Read more… Source: The ...
- WantToCry ransomware remotely encrypts files
May 19, 2026
SophosLabs analysts investigated WantToCry ransomware attacks that involved the threat actors abusing the Server Message Block (SMB) service for initial access and then exfiltrating files to attacker-controlled infrastructure for remote encryption. The detection surface is significantly reduced because WantToCry operates without local malware execution, and there is no post-compromise activity beyond exfiltrating files and rewriting ...
- NGINX Rift attackers waste no time targeting exposed servers
May 18, 2026
Exploit attempts are already hammering a newly disclosed NGINX bug dubbed “NGINX Rift,” proving once again that attackers read patch notes faster than most admins. Researchers at VulnCheck said they are seeing active exploitation tied to CVE-2026-42945, a heap buffer overflow flaw affecting both NGINX Open Source and NGINX Plus that was disclosed last week after apparently sitting ...
- Chaotic Eclipse strikes again with another worrying Windows security flaw
May 18, 2026
Threat actors could escalate privileges and gain SYSTEM access on a fully patched Windows 11 device thanks to an unpatched vulnerability which allegedly should have been fixed years ago, new reports have claimed. A researcher with the alias Chaotic Eclipse recently disclosed a Proof-of-Concept (PoC) exploit for a zero-day vulnerability they named “MiniPlasma”. In a new GitHub entry, ...
- NYC Health + Hospitals says hackers stole medical data affecting at least 1.8m people
May 18, 2026
New York public health provider NYC Health + Hospitals says a months-long data breach that allowed hackers to steal personal data, medical records, and fingerprints scans affects at least 1.8 million people. NYCHHC is the largest public health system in the United States and provides healthcare to over a million New Yorkers, the majority of whom are uninsured or ...
- Scammers are targeting World Cup fans
May 16, 2026
We’re less than a month away from the biggest sporting event of the year, the FIFA World Cup, and scammers are already busy stealing money, passwords, and other sensitive data from fans and visitors, experts have warned. Kaspersky has published a breakdown of the different scam techniques cybercriminals are using to target football fans as they ...