GitHub Repositories Used to Distribute Malware Disguised as AI Tools

The Hacker News reports that cybersecurity researchers have identified nearly 7,600 malicious GitHub repositories created as part of a campaign named FakeGit. The repositories pose as legitimate artificial intelligence tools, skills or Model Context Protocol (MCP) servers, but deliver a malware family known as SmartLoader. The campaign uses convincing developer profiles, copied project documentation and realistic README files to appear legitimate. More than 800 of the repositories specifically target developers and data scientists looking for AI tooling. When users download and execute the code, SmartLoader is deployed, providing attackers with initial access to the victim's system. The scale and sophistication of the campaign suggest that attackers are investing significant effort in making the malicious repositories appear credible and difficult to distinguish from genuine open-source projects.

Why this matters for UK organisations

GitHub is a trusted platform for software development, and many UK organisations encourage or require developers to use open-source libraries, frameworks and tools to accelerate development. The FakeGit campaign exploits that trust by creating repositories that look legitimate and appear to offer useful functionality. For organisations where developers routinely download and integrate code from GitHub, this creates supply chain risk: malicious code can enter the development environment or production systems through what appears to be a legitimate open-source project. The targeting of AI tools is particularly relevant given the current interest in integrating AI capabilities into business applications, data pipelines and automation workflows. Developers looking for AI libraries, model servers or integration tools may be more likely to download and experiment with code from unfamiliar repositories, increasing the risk of compromise. The scale of the campaign, nearly 7,600 repositories, suggests that attackers are using automation to create and maintain a large number of malicious projects, making manual detection difficult.

What to review

Organisations with software development teams should review how developers evaluate and approve third-party code, whether controls exist to scan downloaded repositories for malicious content, and whether development environments are sufficiently isolated from production systems. Clear guidance on verifying repository authenticity, checking contributor history, reviewing commit activity and using trusted sources is a practical step many organisations can take. Development teams should be encouraged to prefer repositories with established maintainers, active communities, transparent development history and verifiable provenance. Where possible, code downloaded from GitHub should be scanned using static analysis tools, malware detection engines or sandbox environments before being integrated into development workflows. Organisations should also consider whether development environments have appropriate network segmentation, endpoint detection and response coverage, and access controls to limit the impact of a compromised developer workstation. For organisations using AI tools or libraries, reviewing the source and trustworthiness of those tools is a sensible governance step, particularly where they are integrated into production systems or handle sensitive data.

Source: The Hacker News