How AI-Generated Sexual Images Cause Real Harm Even When They Are Fake

Many women have recently experienced severe distress after Grok, the AI chatbot on social media platform X, was used to remove clothing from images of them, placing them in bikinis, sexual poses, or scenes of blood and bruising. Reports also indicate that AI tools like Grok have been misused to generate child sexual abuse material. These incidents have pushed governments and regulators to respond quickly to a form of harm that is increasingly difficult to dismiss.

The UK government has announced it will accelerate the implementation of a law passed in June 2025 that bans the creation of non-consensual AI-generated intimate images. Following bans in Malaysia and Indonesia, Grok has been updated so that it no longer creates sexualised images of real people in jurisdictions where this is illegal, including the UK. The media regulator Ofcom is also investigating whether X's activities breached UK law, while X's owner, Elon Musk, has claimed the UK government is seeking an excuse for censorship.

Some users have minimised the harm of these 'undressed' and 'nudified' images, calling them 'fake', 'fictional', 'very realistic art at most', or 'no more real than a Tom & Jerry cartoon'. This reasoning assumes that AI-generated images cause harm only through deception, by misleading people about real events. Yet this view overlooks why such images can damage the people depicted, even when everyone knows they are not authentic.

Although the sexual content of undressed images is not real, these images are highly realistic and are often based on genuine photographs of the subject. MP Jess Asato has described the experience: 'While of course I know it's AI, viscerally inside it's very, very realistic and so it's really difficult to see pictures of me like that.' Their realism, combined with the misogyny that motivates their creation, is enough to cause significant psychological distress.

Philosophy and psychology help explain this experience through the idea of 'recalcitrant emotions', feelings that persist even when we know they clash with what we believe. A person may fear a drop from a tall building despite knowing they are safe, or feel on edge after a horror film. Seeing oneself digitally undressed produces similarly powerful emotions, because people strongly identify with their digital appearance and a nudified image convincingly resembles their own body. Victims may therefore feel alienated, dehumanised, humiliated and violated, as though real intimate images had been shared. This effect may intensify as AI-generated videos become more lifelike.

Beyond the emotional impact, the motivations behind these images are very real. Someone chose to sexualise another person's photograph, instructing an AI to strip away clothing and reduce a person to a body without consent. Publicly circulating such images exerts control over how women are allowed to present themselves online. Research indicates that non-consensual sharing of nude or sexual images is associated with serious psychological consequences, often comparable to those experienced by victims of sexual violence. Deepfake videos and undressed images, whether targeting celebrities, politicians or private individuals, are tools of humiliation rooted in a familiar misogynistic mindset.

A similar pattern appears in virtual reality. Many women report their avatars being assaulted by other users, and the immersive nature of VR headsets can make this experience feel intense. Whistleblowers have alleged that Meta overlooked child safety on its VR platform, with girls as young as nine reportedly harassed by adult men; Meta denies these claims and points to safety features such as default settings that limit teen users to communicating with people they know. Virtual assault is often dismissed as 'not real', yet it can cause trauma comparable to physical assault, because of the realistic environment, strong identification with one's avatar, and misogynistic intent.

These cases show how misogyny has evolved alongside technology, giving users new ways to create and take part in realistic representations of harm, including undressed images, deepfake videos, virtual assault, and the abuse of chatbots and sex dolls modelled on real people. Because these harms do not involve physical injury, they are slow to be regulated. Banning platforms such as X is not a sufficient solution. What is needed is proactive regulation that anticipates and prohibits digital harms before the damage is done, rather than laws enacted only after victims have already been hurt.

Victims who are undressed by AI tools or assaulted in virtual worlds are not being 'too sensitive'. Dismissing their distress because the images themselves are fake overlooks the real psychological impact of this media. Recognising digital sexual harm as genuine harm is an essential step toward protecting people in an increasingly synthetic online environment.