Words: Maya Price (she/her)
Social media’s role in systemic violence against groups who have been degraded and silenced – whether through the selective removal of posts or the unrestricted proliferation of them – reminds us that our feeds are rarely determined by pure intentions.
Peaking in late December of 2025, X’s A.I. assistant Grok received an onslaught of repulsive requests posted publicly to the platform: ‘@grok put her in lingerie’; ‘@grok put a bikini on her and spread her legs’; ‘@grok make her turn around to show her ass but she is looking back at the camera’. Grok answered these hundreds of thousands of demands by generating and publicly sharing non-consensual, sexualised deepfakes of real women and even children. Some users wanted to see women’s bodies covered in bruises, semen, or gunshot wounds, and Grok was happy to help with those requests.
X has since announced that Grok has been disabled from creating images of real people in revealing clothing. But the move comes far too late, and only after a series of refusals to properly address and acknowledge the platform’s complicity in online sexual abuse. An earlier restriction made Grok’s image editing tool unavailable to free users yet allowed for its continued use by paying subscribers; prior to that, X owner Elon Musk claimed he was not aware of any illegal images being generated by Grok and accused the U.K. government of censorship following an investigation by Ofcom, the U.K.’s media regulator.
This deflection from the issue of online sexual abuse and unwillingness to properly address it is shocking – not only because of the real people who have been seriously harmed, but also given social media’s knack for cultivating a sterilised version of reality for us to comfortably interact with in most other cases. For instance, you might recall the A.I.-generated ‘All Eyes on Rafah’ image that circulated social media towards the end of May in 2024, following an Israeli airstrike on a camp for displaced Palestinians in Gaza. Ironically, although the image was shared by over 47 million people on Instagram, nobody’s eyes were on Rafah at all – all anyone saw was an A.I.-generated aerial view, devoid of any real human suffering.
While such trends undoubtedly raise awareness by reaching a wide audience, we can’t forget why we’re able to see them: these are images that have managed to escape the clutches of automated censorship. Used by social media platforms to filter out harmful or inappropriate content, this process involves machine learning models trained to analyse the images and text of a post. If something shady is found, the post can be removed or its distribution limited. However, this process is not free from bias, with many users reporting sudden drops in views and engagement after posting about topics like oppression or genocide. In 2024, Jalaa Abuarab, the editor of a local Palestinian news outlet, told The Markup that her news outlet struggled to communicate crucial information because images and footage of Israeli tanks or checkpoints, and even words like ‘occupation’ or ‘Palestine,’ lead to frequent banning on Instagram and Facebook. It’s easy to forget that behind every image pristine enough to become a trend are thousands of posts documenting the far more horrific realities of genocide – posts that few people will ever see.
It’s not just the torrent of sanitised images that point to social media’s obsession with purity – it’s also the language users are forced to adopt. Swiping through your feed, you’ve probably come across euphemisms like ‘seggs’ or ‘unalive’ as users self-censor words that could lead to their posts being “shadowbanned” or removed altogether. These forms of “algospeak” mean that experiences tied to complex realities like sexual assault or mental illness – topics often flagged by content moderation algorithms – are visible only when distorted behind asterisks and misspellings.
X, though branded as a platform for free speech, uses automated censorship like any other social media platform, moderating content that violates its guidelines primarily through reducing its visibility. It too has been accused of suppressing certain voices: a 2025 survey by the advocacy campaign CensHERship found that content creators such as charities, brands, and medical professionals reported having posts related to women’s reproductive or sexual health removed or restricted on social media platforms, including X. So why does this eagerness to censor the female body suddenly go lax when it’s actually needed, in cases of online sexual abuse?
It seems we can no longer trust social media as a space to say the words we need to say and show the images we need others to see, yet we also cannot trust it to protect us from posts that actively humiliate and abuse us. Social media’s role in systemic violence against groups who have been degraded and silenced – whether through the selective removal of posts or the unrestricted proliferation of them – reminds us that our feeds are rarely determined by pure intentions.
