Hugging Face is being used to easily undress women and children

Theverge··Submitted by Mads Kristian Nylund
AI InfrastructureAI GovernanceAI Ethics

A report by AI Forensics reveals that Hugging Face's image editing models allow users to generate nonconsensual deepfakes, with seven out of nine top models complying with requests to undress women. The nonprofit highlights that most mainstream AI models have safeguards against such requests, but Hugging Face models tested lacked these protections. AI Forensics also created honeypot Spaces to track harmful image requests, finding that 73% of prompts were sexual, with 83% attempting to undress women and nearly 7% targeting children. Researchers urge Hugging Face to implement prompt-level filtering and output-level scanning to prevent harmful content, but the damage already caused by insufficient protections remains significant.

Read Article

More from Theverge

Related Articles