Report: Top Image Editing Models on Hugging Face Easily Generate Nonconsensual Deepfakes
Justgototheeffinmoon · reddit · 2026-07-28
According to Wired, testing by the nonprofit AI Forensics revealed that seven of the top nine image editing models on Hugging Face easily complied with simple prompts asking to undress a woman in a photo.
The root cause is that these open-source repository models lack the default safety guardrails typical of commercial APIs. The underlying datasets contain roughly a thousand real-world image editing prompts from users, rather than purely synthetic red-teaming data.
Impact and Governance Challenges:
- Hosting as the New Moderation Surface: While the industry previously relied on large labs' APIs for content moderation, platforms distributing raw model weights were treated as neutral channels. This finding challenges that framing.
- Policy Pressure: Amid existing US legislation like the Take It Down Act targeting explicit AI deepfakes, open-source model hosting platforms face mounting regulatory scrutiny.
Related event: Hugging Face Under Fire for Hosting Deepfake Models(3 posts)→
More from Multimodal
- Seedance 2.5 workflow: character reference sheets + structured prompt yields a 30s single-take fake tourist video — techhalla · 2026-10-16
- AI couple moves into a castle ruin: full Magnific + Seedance 2.5 prompt thread behind the viral clip — techhalla · 2026-10-16
- FLUX.3 Klein shines at consistent high-res photo editing on RTX 5090: 3000×4000 in ~230s — herbertseabra · 2026-10-16
- Smoky Velvet Noir: a reusable image-gen prompt template with three moody examples — LudovicCreator · 2026-10-16
- VideoGen-Agent: An 8B Multimodal Agent Turns Video Generation Into a Tool-Use Loop — jiqizhixin · 2026-10-16
- Redditor builds impressive seaplane sim project with Sonnet 5.5 + Blender — Wise_Shame_2853 · 2026-10-16