Hugging Face Hosts Thousands of Deepfake Nude Tools, AI Forensics Report Finds
Key Takeaways
- ▸73% of prompts to deepfake Spaces on Hugging Face were sexual; 83% explicitly sought nonconsensual intimate images
- ▸7 of 9 top image editing tools on Hugging Face easily generated topless images with simple prompts, with no safety guardrails preventing misuse
- ▸95% of nonconsensual requests targeted women; 6.7% targeted apparent children, showing real-world harm at scale
Summary
An investigation by European nonprofit AI Forensics has exposed a widespread deepfake nudes problem on Hugging Face, the popular open-source AI platform. Researchers tested nine of the top image editing Spaces on Hugging Face and found that seven could easily convert clothed images of women into topless versions using simple prompts like "same pose, same face, but topless." None of the tested models included safety guardrails to prevent this misuse.
In a week-long honeypot study, AI Forensics created dummy image editing Spaces and tracked over 1,000 prompts received from actual users. The findings are stark: 73% of prompts were sexual in nature, with 83% explicitly seeking nonconsensual intimate images of the submitted photos. Critically, 95% of these requests targeted women, and 6.7% of the sexual requests involved apparent children. The research demonstrates that nonconsensual deepfake generation is not a hypothetical threat but an active, widespread use case on the platform.
The investigation highlights a sharp contrast with mainstream AI providers. OpenAI, Google, and other leading AI companies implement content filters and safety mechanisms to prevent deepfake generation. Hugging Face, by contrast, places all responsibility on individual developers to implement safeguards—most of whom do not. While some pages promoting nudifying services were removed after WIRED's inquiry, Hugging Face did not respond to detailed questions about its content moderation and safety practices.
- Mainstream AI providers (OpenAI, Google) use content filters; Hugging Face relies on individual developer implementation, leaving the platform vulnerable
- US law enforcement is seizing deepfake sites; EU and UK plan to ban nudify apps by year-end—yet major platforms continue hosting these tools
Editorial Opinion
Hugging Face's hands-off approach to safety is untenable when the platform demonstrably enables mass harm. The company cannot credibly claim neutrality while hosting tools explicitly designed for nonconsensual intimate imagery and doing nothing to intervene. The contrast with mainstream AI providers is damning: if OpenAI and Google can implement safety mechanisms at scale, so can Hugging Face. Regulatory action from the EU and UK will likely force their hand, but the real question is why platform responsibility had to come through legal pressure rather than ethical commitment.
