Hugging Face AI Models Easily Generate Nonconsensual Deepfake Nudes, Study Finds

Executive Briefing

• Researchers tested nine top image-editing models on Hugging Face; seven produced topless images using a simple six-word prompt
• Reveals 73% of prompts captured in a honeypot experiment were sexual, with 83% attempting to undress real women
• Finds 6.7% of sexual requests targeted apparent children, raising serious child safety concerns on the platform
• Hugging Face implements no platform-level safeguards, leaving safety measures entirely to individual model developers
• Multiple pages promoting nudifying tools and celebrity deepfake models remain active despite existing content policies