An investigation by the nonprofit AI Forensics has found that Hugging Face, a major open-source AI platform valued in the billions, is widely used to generate nonconsensual nude "deepfake" images. The finding is significant because it shows that even as US, EU and UK authorities move to crack down on "nudify" tools, a large mainstream AI hosting platform is still enabling the technology largely unchecked, with little effort required to bypass any safeguards.

Researchers tested nine popular image-editing tools ("Spaces") on the site and found seven could easily turn a clothed photo of a woman into a topless one using a simple six-word prompt, with no need to circumvent safety measures. In a separate test, decoy Spaces set up by the group logged more than 1,000 user submissions over a week, of which 73% of prompts were sexual, 83% of those sought to undress or sexualise the subject (95% of them women), and 6.7% appeared to target children. Hugging Face did not respond to Wired's questions about its moderation practices, though it does have policies banning child sexual abuse material and non-consensual sexual deepfakes; some flagged pages were removed after Wired's enquiries.