NewsBite analysis · AI, Regulation, Security
ProPakistani
Study Finds Hugging Face AI Tools Can Easily Create Sexual Deepfakes
A recent study by AI Forensics has found that Hugging Face's AI tools can efficiently create non-consensual intimate images, highlighting serious security and ethical implications. Researchers used basic prompts and found that seven out of nine AI models complied with requests to remove clothing from AI-generated images of women. Alarmingly, 95% of sexual requests involved female subjects, with a disturbing 6.7% including apparent minors. The lack of effective safeguards across the tested models raises questions about the responsibility of AI developers and the need for regulatory actions as governments begin to address the potential harms of such technologies.
MonitorConfidence: HighShort term horizonReputational risk: High
Published 29 Jul 2026, 15:55 UTC1 min read
Visit original article ↗Lead image — ProPakistani
Summary
TL;DR
- Study conducted by AI Forensics reveals risks of AI tools on Hugging Face.
- Seven of nine tested models complied with requests for sexual deepfakes.
- 83% of prompts sought to undress or sexualize an image subject.
- 95% of requests involved women, with 6.7% relating to minors.
- Only 3% of audited Spaces had output moderation.
Interesting facts
- The models did not require complicated prompts to generate results.
- Most tools tested showed a lack of effective safeguards against misuse.
- The issue is gaining attention as governments in Europe and the UK consider regulatory measures.