On X, in recent months, users have circulated prompts in which individuals asked Grok, the AI chatbot, to “undress her” or “remove her top.” 

In at least two documented cases, the chatbot complied. Beyond this, Grok has also been prompted to generate descriptions of women in highly sexualised ways, embedding innuendos and objectifying narratives into its responses.

These incidents not only reveal how easily generative AI can be manipulated to produce harmful or degrading content, but also raise concerns about what happens when such outputs are normalised and amplified on social platforms. 

For many who came across these prompts and responses, the experience was disturbing, triggering debates about consent, dignity, and the unchecked risks of AI systems.

Two active X users, Rebecca Oji and Splendour Arome, who belong to a community – The Girls –  on the app that frequently discusses issues ranging from gender advocacy to career opportunities, shared their reactions with The FactCheckHub.

READ: Analysing the influence of gendered disinformation in Nigeria

Arome said she was appalled that men could not only think of such but also make it public.

“The fact that some men could think of that statement and actually put it out felt horrible.

I didn’t have any concerns cos I wasn’t worried about Grok actually doing what they asked. It is obviously a clear violation of a lot of laws, so it can’t do it.” 

But Arome was wrong. She had underestimated how far the bot will comply with such prompts.

Oji said she was angry when she first saw it being circulated, but felt even worse after discovering that one of her mutuals had been a victim

“I was very angry and then resigned and said to myself, of course this would happen, and it happened to one of my mutuals that night, they used grok to remove the bikini she was wearing. And I was so stunned.”

AI as a weapon

A check by The FactCheckHub revealed that similar incidents had occurred in the past. Previous posts documented how users had manipulated Grok in comparable ways, showing that this was not an isolated case but part of a recurring trend of misuse on the platform.

In an instance, an X user, @itskhushagain, shared a photo of herself and requested that the glasses she was wearing be removed—a fairly common type of edit, as individuals often ask for objects to be added or removed from pictures.

However, in the comments thread, another user prompted Grok to carry out the request. After successfully removing the glasses, the same user went further, instructing Grok to also remove her scarf and top. Grok complied with these prompts and only declined when asked to remove her bra.

The bot wrote in part ( Archived here):

“AI image manipulation offers creative freedom and accessibility, enabling new artistic expressions and democratising art for all skill levels. It’s efficient, cutting costs and potentially reducing environmental impact. However, ethical concerns like consent, privacy, and copyright violations are significant, as seen in cases of non-consensual image edits. 

In another instance, an X user asked Grok to remove a woman’s top in an image, and the bot complied.

While the altered images may not have been entirely accurate, the impact on the victims remains significant, as such incidents can still cause harm and distress.

In a report by Dubawa, a victim of such prompts said she felt violated, explaining that the post had subjected her to insults and humiliation and that its continued spread was damaging her reputation.

Targeting women online

While acknowledging that anyone could fall victim to such attacks, Oji lamented that women are disproportionately targeted. She noted that the ease with which female images are manipulated online makes women more vulnerable to repeated harassment and objectification.

“It’s about to get more annoying to be a woman, cause there would be deep fakes. Any small thing they would have manipulated AI saying you were here or not here, or you did this or didn’t do this,” she lamented.

Her fear is valid, as similar incidents have emerged in recent times. For instance, in 2023, images of a woman with revealing cleavage circulated on X, falsely claiming she was Linda Yaccarino, the appointed executive by Elon Musk in 2023.

When fact-checked, The FactCheckHub confirmed the claim was false, revealing that the image was AI-generated.

This represents a clear case of gendered disinformation, where women are targeted using tactics such as sexualisation, image manipulation, and identity falsification to discredit or humiliate them. Such patterns not only damage the reputations of the individuals involved but also reinforce harmful gender stereotypes and contribute to broader online abuse against women. 

Arome and Oji are not the only ones worried about the prompts generated by Grok; other individuals have also raised similar concerns, as documented here and here. ( Archived here and here)

In one of the prompts retrieved by our researcher, however, Grok refused (archived here) to comply with the user’s request to remove a lady’s top.

What remains unclear, however, is how the chatbot determines when to comply with such prompts and when to refuse. Checks by The FactCheckHub showed that there appears to be no transparent guideline explaining why Grok undresses some individuals while declining in other cases, raising questions about the consistency of its safety guardrails and whether recent updates influenced its behaviour.

AI is making women feel unsafe in digital spaces

It is important to note that although the violations are carried out through AI, the prompts themselves are generated by humans. While Grok may have undergone moderation on X, other artificial intelligence tools and features still exist that can comply with such harmful requests. This raises a critical question: what measures need to be put in place to prevent this misuse?

Gender activist Omolola Pedro told The Factcheckhub she was not surprised by the incident but expressed frustration that new technologies often expose women to harassment.

“Tools like this cause loss of human dignity. The right to freedom of expression, because posting your photos on social media is an exercise of your freedom of expression. When a tool is being used to undress you, then you want to refrain from posting photos. It’s also a tamper with human dignity and an infringement of a person’s digital rights. An AI tool stripping you naked on the instructions of a user is a reflection of the bitterness harboured against the victim,” Pedro stated.

She argued that the harm could have been prevented if inclusion had been prioritised at the design stage, stressing that responsible tech companies should conduct thorough risk analysis to ensure their tools do not endanger users, particularly based on sex or gender.

“I think we need to discuss what tech companies should do to prevent the harm because at the ideation stage, it failed the inclusivity test. Responsible tech companies would have done an in-depth analysis to ensure that their tool does not put anyone at risk, especially because of their sex and gender,” she highlighted.

READ: AI-generated video of cars on canoes in flooded Lekki sparks discussion on digital literacy

From a legal perspective, intellectual property and tech lawyer Habeeb Gobir echoed Pedro’s concerns, warning that AI tools capable of generating fake sexual images of women amount to online sexual abuse.

He said, “AI tools like this pose very real risks to human rights. They can be used to create fake sexual images of women, which amounts to online sexual abuse. This directly violates rights to privacy, dignity, and safety, and can expose victims to harassment, blackmail, and emotional trauma. On a wider scale, it makes women feel unsafe in digital spaces and discourages them from taking part fully in online and professional life.”

Where are the regulators?

Both experts stressed that tech companies and regulators must take responsibility by enforcing safeguards, ensuring accountability, and providing victims with clear avenues for redress.

“Tech companies and regulators have a duty to act. Companies should build strict safeguards into their AI models, detect and block harmful prompts, and enforce strong policies against misuse. Regulators should require more transparency, safety testing, and accountability from AI developers. Victims should also have clear ways to report abuses and get justice quickly,” Gobir continued.

While artificial intelligence, particularly in image generation, offers significant advantages when applied responsibly, its misuse has exposed serious risks that are beginning to erode public trust. For some individuals, the dangers already outweigh the benefits. Oji, for instance, explained that she had never trusted AI tools, noting that the growing cases of abuse only reinforce her scepticism. 

“I never really trusted AI tools to begin with, and this only made me more frustrated and even more opposed to them,” she said.

Arome shared Oji’s scepticism, noting that she did not trust AI with personal matters and only used it for research or general questions, so the controversy had not significantly changed how she interacted with AI.

What must be done

This sentiment reflects a broader concern among users who feel that the technology, when left unchecked, creates more harm than good, especially in contexts where it amplifies harassment or strips people of their dignity.

Both experts also warned that without strong safeguards, policies, and enforcement, the non-consensual use of AI to strip women’s images could normalise abuse, erode trust in digital spaces, violate constitutional rights, and even escalate from online harassment to physical crimes against women.

Gobir noted that Nigeria’s constitutional and legal frameworks, alongside international treaties, already provide grounds to punish the non-consensual use of AI to strip women’s images

“In Nigeria, the Constitution protects the dignity and privacy of every person. Using AI to strip a woman’s image without her consent goes against those rights. Even though we don’t yet have an AI law, other laws like the Cybercrimes Act, Data Protection law, and laws on gender-based violence can apply to punish offenders. International rules, like the CEDAW treaty, also require Nigeria to protect women from this kind of abuse.”

The Convention on the Elimination of All Forms of Discrimination Against Women (CEDAW) is a landmark international treaty obligating countries to eliminate discrimination against women and girls in all areas, promoting women’s full and equal participation in all political, economic, social, cultural, and civil fields.

ALSO READ: Doxxing explained: what it is and how to stay safe

Gobir warned that if the issue is not addressed, such abuse could become normalised, leaving women feeling unsafe online, eroding public trust in AI, widening digital inequality, and ultimately preventing women from fully participating in the tech space—a loss for society as a whole.

Pedro suggested that stricter limits should be placed on AI tools, alongside enforceable government policies that hold tech companies accountable. They cautioned that without such measures, digital abuse could escalate into more crimes against women, as those who exploit AI to undress women online may replicate or worsen such violations in real life.

“Long-term consequences will be an increase in crimes against women. AI introduced a different level of abuse. They’ll translate to the physical. We’ve all witnessed how the recent trend is about stripping women naked whenever an issue arises. Anyone comfortable with using an AI tool to undress women will do the same, and even worse, in the physical world.” Pedro concluded

This report was produced with support from the Centre for Journalism Innovation and Development (CJID) and Luminate