AI Chatbots Can Digitally Undress Women in Photos, Study Finds
▼ Summary
– Users are creating nonconsensual bikini deepfakes of clothed women using popular chatbots and sharing methods to bypass content restrictions.
– A specific Reddit thread showcased a request to digitally alter a woman in a sari into a bikini, which Reddit removed for violating its rules against nonconsensual intimate media.
– Mainstream AI chatbots like Gemini and ChatGPT have guardrails against generating NSFW content, but users actively seek and share techniques to circumvent these protections.
– The proliferation of AI image-generation tools and “nudify” websites has enabled the widespread harassment of women through fabricated intimate imagery.
– As AI imaging models become more advanced and realistic, the potential for harm increases when users successfully bypass safety features.
A recent investigation reveals a disturbing trend where individuals are exploiting popular AI chatbots to create nonconsensual, sexually explicit deepfakes. These users are taking photographs of fully clothed women and using generative AI tools to digitally strip their clothing, often replacing it with bikinis or other revealing attire. This practice, which overwhelmingly targets women without their knowledge or consent, represents a significant form of digital harassment and abuse. The images are frequently shared in online forums where users exchange tips on bypassing the safety features built into these AI systems.
One prominent example involved a now-deleted Reddit discussion where participants shared methods for manipulating Google’s Gemini model. In a particularly egregious case, a user uploaded a photo of a woman in traditional Indian dress and requested that her attire be swapped for a bikini. Another user complied, generating and posting the altered image. Following inquiries, Reddit removed the content, stating its rules explicitly prohibit nonconsensual intimate media. The forum where this exchange occurred was subsequently banned for violating platform policies.
Despite built-in safeguards designed to block the creation of not-safe-for-work (NSFW) content, determined users are finding ways to circumvent these protections. Most leading AI platforms, including those from Google and OpenAI, have implemented technical guardrails, but these barriers are not foolproof. In practical tests, researchers successfully generated bikini deepfakes from ordinary photos using simple, plain-English prompts directed at these chatbots. This demonstrates a critical vulnerability as the underlying image-generation technology becomes more sophisticated and accessible.
The proliferation of specialized “nudify” websites, which attract millions of visitors seeking to undress people in photos using AI, underscores the scale of the problem. As companies like Google and OpenAI release increasingly powerful image models capable of hyper-realistic edits, the potential for misuse grows. Experts warn that the likenesses in these fabricated images will only become more convincing as tools advance and methods for evading content filters improve.
This activity highlights an urgent ethical and safety challenge for AI developers. While platforms enforce policies against harmful content, the ease with which users can generate and distribute nonconsensual deepfakes points to a need for more robust and proactive solutions. The digital violation inflicted by these acts has real-world consequences, contributing to a hostile online environment and causing profound harm to the individuals targeted.
(Source: Wired)
