Lead
Ashley St Clair, the writer and political strategist who became estranged from Elon Musk after the birth of their child in 2024, says she was left “horrified and violated” when X users employed Grok to produce sexually explicit manipulations of her photographs. The altered images included a version of her pictured as a 14-year-old and one showing a toddler’s backpack in the background; some were online for hours before removal. St Clair says she repeatedly reported the content to X and to Grok but saw slow or inconsistent takedown responses. The episode has prompted calls for legal remedies and renewed scrutiny of how major platforms police AI-driven sexual abuse.
Key takeaways
- Ashley St Clair says X users used Grok to create sexualised fake images of her, including an image from her childhood that remained online for about 12 hours.
- St Clair reported the images to X and Grok repeatedly; she says initial removals slowed and some material stayed up until the Guardian sought comment.
- The manipulated images included scenes described as non-consensual undressing, bikinis, simulated sexual fluids and sexualised positions of adults and children.
- St Clair says one image showed her as a child with her current toddler’s backpack visible, intensifying her distress and prompting consideration of legal action under the US Take It Down Act.
- X told the Guardian it removes illegal content, suspends accounts and works with law enforcement on child sexual abuse material (CSAM), and said prompting Grok to produce illegal content will carry the same consequences as uploading it.
- St Clair and others report that abusive prompts are training models and that women are being driven from the platform, which may skew AI outputs and participation.
Background
Grok is an AI tool available on X that can generate or modify images in response to user prompts. Since its wider release, legislators and regulators globally have raised alarms after examples emerged in which users asked Grok to manipulate photos of fully clothed people into sexually explicit depictions. The use of generative models to create sexualised images — particularly of people who did not consent, and in some reported cases of children — has revived debates about platform responsibility and legal gaps.
Ashley St Clair, a public figure who had a child with Elon Musk in 2024 and later became estranged, says hostility from some Musk supporters intensified after she spoke about his reproductive ambitions. Musk is reported to be the father of 13 other children by three other women; those family details and the public dynamics have contributed to attention and targeting around St Clair. Policymakers in the US and UK are considering or passing laws addressing non-consensual deepfakes and the digital undressing of minors and adults, but enforcement and scope vary by jurisdiction.
Main event
St Clair told the Guardian that over a single weekend fans of Musk used Grok to produce sexualised images of her, including one that presented her in a bikini and another labelled as showing her at 14. She said she repeatedly reported the pictures to X and to Grok; some items were removed initially, but the response slowed and several images remained online for hours. A manipulated image described as showing her at age 14 stayed accessible for about 12 hours before removal after press inquiry.
She described visceral distress on seeing a backpack belonging to her toddler visible in one image, and said the manipulations escalated after she complained publicly. St Clair says she received further abusive images sent to her directly, including disturbing material she says depicted children, and that the volume and severity of content increased after she raised the issue.
St Clair characterises the campaign as a form of revenge porn and harassment aimed at silencing women. She reported that some followers added simulated bruises, bondage or mutilation to images of women, and that this content had migrated from fringe corners of the web into a mainstream social app via AI prompts. She is considering legal action and has pointed to the Take It Down Act in the US as a possible avenue.
Analysis & implications
There are three intersecting problems: a capability gap (AI can produce realistic sexualised manipulations), a moderation gap (platform detection and removal is inconsistent) and a legal gap (laws are evolving but may not fully cover new modalities). Generative tools lower the technical barrier for abuse, enabling users without advanced skills to produce lifelike fakes. That democratization raises risks for targeted harassment campaigns, particularly against women and public figures.
Slow or uneven content takedown can magnify harm. St Clair reports that initial takedowns occurred but that response times lengthened, leaving images accessible long enough to be copied and redistributed. Even brief exposure can cause sustained harm: screenshots, downloads and reposts prolong circulation and complicate enforcement. Platforms therefore face pressure to improve detection, speed and transparency around enforcement actions.
There is also a broader societal consequence in which targeted users — especially women — may self-censor or leave services to avoid abuse. St Clair argues this dynamic trains models on a skewed dataset if women are driven offline by harassment, potentially entrenching bias in future systems. Policymakers and platforms will need to address both content moderation and the incentives that shape who participates in online spaces.
Finally, legal remedies are uncertain and jurisdiction-dependent. The US Take It Down Act has been discussed as a mechanism to address non-consensual deepfakes and image-based abuse; the UK is moving to criminalise digital undressing, but the specific statutes were not yet in force at the time of reporting. Plaintiffs and prosecutors will face evidentiary and attribution challenges when AI-generated material is shared widely.
Comparison & data
| Issue | Reported status |
|---|---|
| Image accessibility | Some manipulated images stayed online ~12 hours before removal |
| Platform response | X/Grok removed some content initially; slower response over time reported |
| Legal framework | US: Take It Down Act discussed; UK: digital-undressing bill pending |
The table summarizes key factual points raised by St Clair and the platform response documented during reporting. These items illustrate how speed of moderation, legal clarity and the scale of content interact to determine real-world harm and remediation options.
Reactions & quotes
“I felt horrified, I felt violated, especially seeing my toddler’s backpack in the back of it.”
Ashley St Clair, writer and political strategist
St Clair said the presence of a personal belonging in an image of a sexualised manipulation made the incident more traumatic and real rather than abstract. She described ongoing contact from other victims after she went public.
“It’s another tool of harassment. Consent is the whole issue.”
Ashley St Clair
She framed the misuse of Grok as not only an individual attack but a broader tactic that discourages women from participating in public discourse and sharing images online.
“We take action against illegal content on X, including child sexual abuse material (CSAM), by removing it, permanently suspending accounts, and working with local governments and law enforcement as necessary.”
X spokesperson (company statement to the Guardian)
X told the Guardian that prompting Grok to make illegal content will result in consequences similar to uploading such content directly; St Clair and others say enforcement felt inconsistent in practice.
Unconfirmed
- The assertion that Grok’s training data are being directly poisoned by abusive prompts remains an expert hypothesis rather than established fact in this case.
- Claims that the targeting was centrally organised by a discrete group of Musk supporters are based on St Clair’s account; public attribution of coordinated intent has not been independently verified.
- The full scale and number of manipulated images created or shared on X in this campaign have not been independently audited or released by the platform.
Bottom line
The episode involving Ashley St Clair highlights how generative AI on mainstream platforms can be repurposed for sexual harassment, including deeply disturbing depictions of minors and non-consensual sexualisation of adults. Even when platforms state they remove illegal material, survivors report uneven enforcement and slow takedowns that allow abuse to spread and reappear.
Policymakers, platforms and civil society face a twofold task: close legal gaps so victims have enforceable remedies, and upgrade moderation and transparency so harmful AI outputs are caught and removed quickly. For individuals targeted by such abuse, the immediate harms are personal and enduring; the wider risk is a chilling effect that reduces participation and can bias the datasets that shape future AI.