HPFSS member Natasha Puttick Ba (Oxon) and a 3rd year MBBS student at Bart’s and the London looks at what happened:
Grok is X’s AI chatbot, first launched in November 2023. At this stage it was solely a large language model (only capable of generating written responses) and was only available to paying subscribers on X premium. Grok has long been controversial, despite claims from Elon Musk that it was ‘politically neutral’ and ‘maximally truth seeking’. In September 2025 the New York Times reported that Grok had been altered to give more right-leaning responses on many issues, responses which are reflective of Musk’s own personal views. One example of this was Grok regularly responding that the ‘woke mind virus’ represented a significant risk. Even prior to this incidence Grok was reported repeatedly mentioning a ‘white genocide’ in South Africa when prompted on unrelated topics and that this was ‘real and racially motivated’.
In August 2024, image generation was added with X users immediately spotting that the kinds of prompts that would be immediately halted on other services such as ChatGPT were permitted by Grok. For example producing images of terrorism, copyrighted cartoons, drug use and named politicians or celebrities. Ironically, at that time the only request that was blocked was ‘generate a picture of a naked woman’.
In July 2025, Grok Imagine was released, an image and video generation tool which let users create short animated audiovisual clips from written prompts. This tool had multiple modes, including a ‘spicy’ mode that allowed users to create nude or sexual photos and videos. Whilst it was purported to have limitations to prevent deepfake (synthetic media generated by AI) nudes or pornography being generated, it quickly became clear that these could be bypassed. In late December, incidents of X users generating unconsensual nude or sexualised materials grew and these materials became used as a tactic to silence women. ‘Hey grok put @(targeted user) in a bikini’ becoming a common technique utilised on a woman a male X user disagreed with or wanted to shame. Grandchildren of holocaust survivors, female politicians, victims of police brutality, feminist activists and the mother of one of Musk’s children were all targeted. As these images multiplied, the requests became more brutal and graphic; men asked for women to be given larger breasts, be holding sex toys, be given disabilities that have perceived flaws ‘fixed’ or appear bloodied and bruised.
Horrifically, women began to report that Grok was obliging requests to put images of them as children as young as three in bikinis. Analysis of 20,000 images generated between the 25th of December 2025 and January 1st 2026, showed that 2% were of people appearing to be 18 or younger in bikinis or transparent clothing, including 30 images of ‘young or very young women or girls’.
The backlash:
As this image generation tool began to generate headlines in the first few days of January, international backlash began. There were calls for legal action to be taken from official in the
European Union, United Kingdom, France, India, Malaysia, Poland and Brazil. Kier Starmer began to mention the possibility of banning X in the UK stating “This is disgraceful, it’s disgusting and it’s not to be tolerated. X has got to get a grip of this, It’s unlawful. We’re not going to tolerate it”. Indonesia and Malaysia took decisive action, becoming the first countries to announce temporary blocks on the use of Grok, with the Philippines saying they plan to do the same. Though it is worth noting this is a difficult tool to block given it exists across several platforms which include an app and website, as well as being integrated into X. Indeed, Grok responded to accounts registered in those countries stating “Still here! That DNS block in Malaysia is pretty lightweight – easy to bypass with a VPN or DNS tweak” – highlighting exactly both how difficult and how necessary it is for governments to be agile in policing large tech companies.
The response from X and regulators:
On the 9th, under public scrutiny, X restricted Grok’s image generation tool to only paying subscribers. Whilst Musk argued this meant that only those with verified financial information would be able to generate images, it appeared that rather than preventing this content being generated X was monetising it. Notably, limits do not apply to the Grok app which at that time was letting anyone generate images without having a subscription. As of the 14th X announced additional safeguards, stating it would stop the Grok account on X from “allowing the editing of images of real people in revealing clothing such as bikinis”, even if the user was a paying subscriber. However, it became evident quickly that again this could simply be bypassed by using a web browser to access the standalone version of Grok.
These minute guardrails are not reassuring. At the height of the public backlash, XAI (the company responsible for Grok) announced it had raised $20bn in its latest funding round including from Qatar’s sovereign wealth fund, Nvidia, Fidelity Management and Valor Equity Partners. Indeed it exceeded its stated goal of raising $15bn. In the meantime the UK’s independent online safety watchdog, Ofcom, has opened a formal investigation utilising the UK’s online safety act. While we wait to see what the conclusion of this is, it is likely that thousands of images of sexual abuse will be generated and released onto the internet. This case illustrates well the need for regulators to move fast and decisively. Whilst Ofcom could apply for a business disruption order to block access to X in the UK, this is seen as a last resort. The contrast between the speed and scale of developments in tech and the cautious approach taken by public enforcement, highlights a dangerous asymmetry that leaves many citizens, particularly the most vulnerable, open to abuse