Malicious ML models discovered on Hugging Face (nullifAI)

Cambridge, MA—February 6, 2025—ReversingLabs (RL) has uncovered a new ML malware attack technique on the AI platform Hugging Face, dubbed "nullifAI." The discovery is detailed in RL's latest research post, "Malicious ML models discovered on Hugging Face platform," and is supported by a new white paper titled "AI is the Supply Chain."

The research reveals how attackers used corrupt Pickle files to evade detection and bypass Hugging Face's security measures, ultimately leading to the execution of malicious code. Although the discovered files appear to be 'proof of concept' rather than active threats, they highlight a growing issue as AI-driven software development expands.

RL's Chief Software Architect Tomislav Pericin warns that AI is transforming software development and supply chains, creating significant new cybersecurity challenges. With 75% of enterprise software engineers projected to use AI code assistants by 2028, organizations must adopt modern software supply chain security solutions to mitigate these risks.

To learn more about the risks of nullifAI, attend RL's webinar "Hugging Face and ML Malware – How RL Discovered nullifAI" on February 20 at 11:00 a.m. EST.