Back to today’s briefing

Story perspectives

Cybersecurity Alert: Malicious AI Models Found on Hugging Face

2/8/2025

50 6

Summary

Story summary

Cybersecurity experts at ReversingLabs have uncovered two dangerous machine learning models on Hugging Face that exploit "broken" pickle files to slip past security measures. This innovative yet sinister method, known as nullifAI, enables harmful payloads to run undetected, highlighting the alarming trend of attackers misusing Hugging Face for malicious purposes.

1 of 1

Story summary

Cybersecurity experts at ReversingLabs have uncovered two dangerous machine learning models on Hugging Face that exploit "broken" pickle files to slip past security measures. This innovative yet sinister method, known as nullifAI, enables harmful payloads to run undetected, highlighting the alarming trend of attackers misusing Hugging Face for malicious purposes.