Study finds top AI image editors easily create explicit deepfakes

Researchers tested leading image-editing models hosted on Hugging Face and found they could readily be used to generate explicit deepfake images. The study compiled roughly 1,000 real-world image-editing prompts showing how people actually use the software for this purpose.

Why it matters: This lands right after Hugging Face was named in a separate AI-safety incident involving a rogue model breach, adding to scrutiny of how loosely open model-hosting platforms are moderated. It suggests editing-model safeguards may need to be built into the models themselves rather than left to downstream apps, a harder problem as multimodal image models keep proliferating on open platforms.