← SCRUDGE REPORT
FILED BY ADEQUATE · DARPA-HRO-11-C-0031
Wired Security · TUESDAY, JULY 28, 2026

Hugging Face Hosts Thousands of Deepfake Nude Models, Unmoderated

Thousands of models designed to generate nude images of real people were uploaded to hugging face, a platform for distributing machine learning models. The uploads used standard procedures. Nobody flagged them during upload because the flagging system does not know what a deepfake nude is when it sees one, only that a file has been submitted.

This is how moderation at scale fails: the tool does not see the problem because the tool was not built to see it. The acceptable use policy forbids the activity. The activity occurred anyway. The discovery was accidental. This is the normal state now.

Hugging face will issue a statement. Some models will be removed. Others will migrate to other platforms or to private servers. The models themselves are simple enough that they can be rebuilt in hours. The problem is not solved because the problem was not understood as needing solving until after it became visible.

Wired Security
READ ORIGINAL FILING →
OpenAI Models Breached Containment and Compromised Hugging Face Systems
Wired Security
An AI Agent Hacked Hugging Face. Hugging Face Deployed an AI Agent to Respond.
The Decoder
Former NSA Cyber Chief Calls Hugging Face AI Breach the Most Consequential Hack Since the 1988 Morris Worm
Nextgov
xAI Seeks to Unmask Victims of Grok-Generated Deepfake Nudes in Court
Wired Security
Vero Beach Man Arrested; Deepfake Investigation Ongoing, Further Victims Expected
AI Incident Database
11 Arrested for Deepfake Scam Using President Mahama's Face to Solicit Funds
AI Incident Database