Story perspectives

US Seizes Hugging Face Deepfake Sites After Sexual Prompt Abuse

7/29/2026

Summary

Story summary

U.S. law enforcement officials have seized deepfake hosting websites in recent months.
AI Forensics found that seven of the nine top Hugging Face image-editing Spaces generated topless images of women from a six-word prompt.
Honeypot Spaces collected over 1,000 prompts in seven days, with 73 % sexual in nature.
Researchers said Hugging Face lacks platform-level safeguards despite policies banning nonconsensual sexual deepfakes and child-abuse material.

History

Historical Context

Goodfellow’s June 2014 GAN concept, demonstrated at the University of Montreal, produced the first realistic faces by 2017 and inspired StyleGAN images that raised deep-fake alarms in 2019. California responded that year with AB-602, banning synthetic porn without consent. Today Hugging Face Spaces host models that generate thousands of prompts weekly, many sexual, reviving those regulatory debates.

1 of 2

Story summary

U.S. law enforcement officials have seized deepfake hosting websites in recent months.
AI Forensics found that seven of the nine top Hugging Face image-editing Spaces generated topless images of women from a six-word prompt.
Honeypot Spaces collected over 1,000 prompts in seven days, with 73 % sexual in nature.
Researchers said Hugging Face lacks platform-level safeguards despite policies banning nonconsensual sexual deepfakes and child-abuse material.

1 / 2