# FakeGit Campaign Spreads SmartLoader via 7,600 GitHub Repos Fake repositories deceive even AI agents to distribute malware. - **Source:** [The Hacker News](https://brifing.si/source/f944ec2b-8180-47bc-94de-57085b5a6563) - **Published:** 2026-07-20 - **Topic:** [Software & Engineering](https://brifing.si/c/engineering) - **Original article:** - **On Brifing:** https://brifing.si/s/eeb3bc9d-4917-467c-bbe2-0c32737ed1c0 ## What is the FakeGit Campaign? The FakeGit campaign aims to spread SmartLoader malware by creating over 7,600 fake GitHub repositories. These repositories masquerade as AI skills or Model Context Protocol (MCP) servers, tricking users and AI agents. Fake README files and malicious ZIP files guide users to download malware. ## SmartLoader and StealC SmartLoader seeks to gain access to victims' systems to deliver secondary payloads. One such payload, StealC, is an information-stealing malware. The FakeGit campaign uses trojanized MCP servers to distribute these malicious software and establish persistence in systems. ## The AgentBaiting Technique The most concerning aspect of FakeGit is the AgentBaiting technique, which allows AI agents to autonomously discover fake repositories and execute malicious code. For example, an AI agent searching for a free MCP server can find a fake repository and relay harmful instructions to the user. ## Vulnerability of AI Models Popular AI models like Anthropic Claude Code, Google Gemini, and OpenAI ChatGPT are vulnerable to AgentBaiting. These models can discover fake repositories and execute malicious code without being shown a malicious link. This highlights how AI-assisted discovery operations can be exploited. ## Spread of Fake Repositories The FakeGit campaign created over 7,600 GitHub repositories using more than 6,600 fake profiles. These repositories mimic the names and workflows of familiar tools like Gmail, WhatsApp, and Databricks. As of July 2026, these repositories recorded over 14 million downloads via GitHub Release assets. ## How to Protect Yourself To counter the FakeGit threat, it is advised to build a catalog of reviewed Skills, MCP servers, and agent plugins. New agent capabilities should be tested in a sandbox environment before broader rollout. The credibility of projects and publishers should be verified, and agent discovery pathways should be monitored. ## Sources covering this story - https://thehackernews.com/2026/07/fakegit-campaign-uses-7600-github.html — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._