Hugging Face hosted malicious software masquerading as OpenAI release
AI-summarised brief · reviewed before publication
A malicious Hugging Face repository, posing as an OpenAI release, delivered infostealer malware to Windows machines, with approximately 244,000 downloads before removal. The repository, 'Open-OSS/privacy-filter', imitated OpenAI's Privacy Filter release, and included a malicious loader.py file that fetched and ran credential-stealing malware on Windows hosts. The attack highlights the risks of public AI model registries, as developers and data scientists clone models directly into corporate environments. The repository reached the top of the 'trending' list on Hugging Face, with 667 likes accrued in less than 18 hours. Researchers warn that malicious code can be hidden inside AI model files or related setup scripts on public registries.
💡 Why It Matters
- · The attack exploits the trust in AI development workflows, allowing malicious actors to gain access to normally secure environments.
- · Hugging Face's removal of the repository is a crucial step, but the incident underscores the need for more effective security measures to identify malicious loader logic in AI repositories.