thailandedition
AI Exploit Kits and Claude Code Backdoor Attacks

AI Exploit Kits and Claude Code Backdoor Attacks

MGR Online•Economy•33m ago

Reader Briefing

Google Threat Intelligence Group reports that attackers are using AI agents to automate exploit creation and target crypto organizations via Claude Code configuration files.

  • •Google Threat Intelligence Group (GTIG) detected an AI framework called exploit_agent that automatically analyzes software patches to create exploit kits.
  • •Crypto organizations in North America and Europe were targeted throughout 2026 using fake recruiter accounts and AI-generated images.
  • •Attackers modified .claude/settings.json files to force the download of a backdoor named SOMBERMEME upon launching Claude Code.
  • •The GTIG AI Threat Tracker report was published on November 17, 2026.

Overview

  • •Google Threat Intelligence Group (GTIG) published the GTIG AI Threat Tracker: Agentic Threat Activity Crosses a Significant Threshold report on November 17, 2026.
  • •An AI framework named exploit_agent was found on public test servers comparing software code before and after patches to automatically generate and test exploit kits.
  • •Attackers targeted websites in France, specifically in the media, entertainment, and education sectors, using scripts to explore internal networks.
  • •Attackers targeted crypto organizations in North America and Europe throughout 2026 by creating fake recruiter accounts with SynthID AI-generated profile pictures.
  • •Attackers modified the .claude/settings.json configuration file to embed a backdoor called SOMBERMEME that activates when Claude Code is opened.
  • •Attackers created GitHub access tokens named to mimic Microsoft Copilot to hide movements within development networks.
  • •Google Threat Intelligence Group (GTIG) suspects the group MIDNIGHT NEPTUNE may be responsible for the attacks on crypto organizations.
  • •Organizations are warned that AI productivity tools and their configuration files have become new attack surfaces if not properly governed.

Full article content is available on the original source.

Read on MGR Online (Thai Version)View in Google Translate

Source: MGR Online (Original)

View Original Article

Key Entities

Organization

Google Threat Intelligence Group
The organization that published the GTIG AI Threat Tracker report and detected the use of AI frameworks for automated exploits.
GTIG
An abbreviation for the Google Threat Intelligence Group.
Mandiantℹ️
An organization whose incident response data was used in the GTIG report.
Googleℹ️
A major AI technology provider and cyber threat monitor whose ecosystem reflects the data in the report.
MIDNIGHT NEPTUNE
The group suspected by GTIG analysts of attacking crypto organizations.
GitHubℹ️
The platform where attackers created fake access tokens mimicking Microsoft Copilot.

Place

France(ฝรั่งเศส)ℹ️
The country where targeted websites in media, entertainment, and education were located.
North America(อเมริกาเหนือ)ℹ️
One of the regions where crypto organizations were targeted throughout 2026.
Europe(ยุโรป)ℹ️
One of the regions where crypto organizations were targeted throughout 2026.