The boys undressing ladies with X’s Grok

Each Thursday, Delve Into AI will present nuanced insights on how the continent’s AI trajectory is shaping up. On this column, we study how AI influences tradition, coverage, companies, and vice versa. Learn to get smarter in regards to the individuals, initiatives, and questions shaping Africa’s AI future.

On June 27, 2025, Nigerian content material creator Asherkine posted a seemingly harmless video asking a younger girl out on a date after she revealed she was single. Shortly after the put up went viral, an nameless Snapchat person generally known as ‘Kenny’ started spreading a fabricated narrative: that the girl within the video, later recognized as Ifeme Rebecca Yahoma, a College of Nigeria Nsukka scholar, was his girlfriend. He created deepfakes, manipulating photos she’d shared on-line to assist the lie. In her picture, Yahoma’s head is barely tilted, her lips pouted. Within the black-and-white picture ‘Kenny’ generates, he has doctored himself to look as if pecking Yahoma. An emoji covers his face. Kennedy gained tens of hundreds of followers off the lie, even promoting Snapchat adverts as his virality surged.

This incident isn’t an outlier; it’s the tip of a rising digital iceberg. Throughout Nigeria and different African nations, on-line customers are using AI instruments like X’s Grok to govern, sexualise, or humiliate ladies. What was crude Photoshop jobs at the moment are photorealistic deepfakes created from sexually specific prompts fed into AI techniques that produce disturbingly life-like outcomes.

In the event that they received’t ship nudes, create them 

Earlier this 12 months, an X person had prompted Grok to undress Nigerian actress Kehinde Bankole. The put up, which circulated broadly earlier than being deleted, despatched shockwaves by means of her fanbase and different involved X customers. If a public determine like Bankole isn’t exempt, on a regular basis ladies have even much less safety.

Gbenga, a self-identified psychological well being guide who made such prompts, later posted an apology thread on X saying: “I really feel a profound sense of regret for not dealing with the state of affairs with the care it warranted.”

Remorseful or not, a rising variety of customers, typically male, have discovered, of their harassment of ladies on-line, a keen participant in X’s AI device which has lately been accused of producing antisemitic commentary amongst different gaffes. These incidents are leaving their victims with not solely emotional harm and reputational smears however little or no authorized recourse.

Are there authorized protections that apply?

When requested if new legal guidelines are essential to deal with how a quickly evolving know-how like AI is enabling abuse in manners beforehand unseen, authorized coverage analyst Sam Eleanya says it should be approached with extra nuance. “The premise that we’d like legal guidelines to manage each new development in know-how just isn’t all the time sound,” he stated. “The higher query is: how does this new device match into the prevailing authorized order?”

In keeping with Eleanya, utilizing instruments like Grok to painting a lady in sexual or compromising methods with out her consent might quantity to prison defamation of character underneath sections 373–375 of the Prison Code Act. If cash is demanded or intimidation is concerned, blackmail, extortion, or conspiracy to take action, costs might observe underneath Part 408 of the Prison Code Act. The Cybercrimes Act of 2015 additionally covers acts like cyberstalking, identification theft, and image-based abuse, all of which might apply to those new situations.

However holding perpetrators to account stays difficult in varied methods, in response to  Queen-Esther Ifunanya Emma-Egbumokei, a company lawyer specialising in worldwide business regulation and the artistic economic system.

Prosecution is usually difficult by jurisdictional ambiguities—the place precisely the perpetrator is supposed to be tried stays unclear. Customers typically function anonymously or cover behind pseudonymous accounts, making it troublesome for identities to be verified and authorized processes to be initiated.

Moreover, Emma-Egbumokei notes that Nigeria, like many African nations, lacks AI-specific rules that would clearly outline and criminalise the misuse of generative instruments for harassment or defamation. Within the absence of such focused authorized frameworks, enforcement is additional weakened, and victims are left weak inside a gray space of the regulation.

Hidden value

An increasing number of Nigerian ladies at the moment are rethinking their presence on social platforms frightened that customers can modify their images inappropriately. “Not posting my photos on right here earlier than they use AI [to] take away my hijab,” wrote X person @wluvnana. 

Whereas some customers on the platform discovered humour within the put up, she was expressing critical concern over a development that she had begun to note. 

“The one factor that appears viable is to take myself out of such conditions, which is absurd. I ought to be capable of put up photos on X if I need to, however now, I don’t have the facility to try this anymore,” she later instructed TechCabal. 

Past doctored images and movies, harassment utilizing generative AI also can seem in text-based codecs, explains Chioma Agwuegbo, govt director at TechHerNG, an organisation targeted on supporting ladies by means of digital literacy and inclusion. 

“There are platforms on-line the place you may go and simulate WhatsApp conversations, simulate Snapchat, Tiktok,” Agwuegbo says. “The phrase ‘generative’ implies that it’s the creation of content material, of media.” 

Compounded by a tradition of disgrace round intercourse and sexuality, and an rising incapability to evaluate the accuracy of media shared on-line, it doesn’t take a lot to trigger lasting harm. “That’s why the best factor to say a few lady to carry her down is ‘I had intercourse along with her’, or ‘I dated her and dumped her’,” Agwuebo says, and AI instruments are making it more and more potential to physician these narratives believably. 

What are the platforms doing?

A part of the difficulty is not only that generative AI might be misused. One other basic drawback is that there are not any appropriate mechanisms in place to detect the misuse that happens significantly within the Nigerian context. 

X used to have a stronger Belief & Security group devoted to content material moderation. Nevertheless, since its October, 2022 acquisition, the headcount for the Belief & Security group has fallen by roughly 30%, in response to Australia’s eSafety Fee. Presently, there isn’t any regional workplace in Africa devoted to regionally conscious content material moderation efforts. The workplace opened in 2021 to serve an African viewers, nevertheless it closed shortly after Elon Musk took over. 

These platforms are equally turning to AI to do the heavy lifting of moderation. It’s cheaper and generally quicker to let automated instruments deal with moderation than to rent a big group of human reviewers. Nevertheless, these instruments will generally battle with nuance and accuracy, particularly when they’re skilled with restricted knowledge for particular areas and native or creole languages. 

When X customers started tagging Grok to generate specific photographs of Nigerian ladies by prompting it to ‘flip her round’, ‘present her bottom’ or ‘take away her garments’, some tried to struggle again by mass reporting the accounts. However time and again, they had been met with an identical chilly response: “This account doesn’t violate neighborhood pointers”.

“I saved asking: if this doesn’t violate your neighborhood pointers, then what does? Their platforms can’t recognise gendered harms which might be peculiar to our Nigerian context, our memes, our slangs which might be utilized by ‘banger boys’ on X. There’s a very large hole in content material moderation,” says Jessica Eni, a coverage affiliate at TechSocietal, who participated in mass reporting efforts.

What different customers  can do

Whereas it is likely to be troublesome to manage AI-assisted sexual harassment on social platforms like X, there are steps daily social media customers can take as a substitute of being unconcerned passersby. That is what customers like Vivian Nnabue, an Ibadan-based social media affiliate, consider. 

When Nnabue noticed posts prompting Grok to undress ladies on X, she made a LinkedIn put up accompanied with screenshots and tags of the perpetrators’ accounts, demanding  that they  be held accountable. 

However she  didn’t cease at a LinkedIn put up. She says she reported the perpetrators to their bosses and, for one who was in class, she despatched experiences to the establishment. Nobody responded. 

What adopted as a substitute, was the perpetrators reaching out to her in her DMs. One in every of them, Gbenga, apologised for what he had carried out including that  he nearly misplaced a contract when an employer got here throughout Nnabue’s LinkedIn put up throughout a background examine. “My inclination was not nudity. [I]  simply need to flex my tech intelligence,” Gbenga wrote to Nnabue. 

Past flagging

Even when platforms tackle their blind spots, advocates like Eni fear that Nigeria’s legal guidelines nonetheless lag. The nation has no single, sturdy framework that totally protects individuals from these evolving types of gendered harassment. Present legal guidelines, such because the Knowledge Safety Act and the Cybercrimes Act, are restricted in scope. There’s additionally a rising concern amongst advocates like Agwuegbo that these legal guidelines might be doubtlessly misused to justify the suppression of free expression or dissent. 

“There’s a Cybercrime Act, sure, however it’s imprecise and heavy-handed. As a result of it’s so ambiguous, it may be twisted nonetheless somebody in energy needs, so long as they’ve cash,” Agwuegbo says. Presently, the Cybercrime Act is getting used to justify fits made in opposition to Senator Natasha Akpoti, a feminine senator who spoke in opposition to sexual harassment by the Senate President. In September 2023, the identical Act was additionally used in opposition to a lady who left unfavourable opinions of a can of tomato paste by Nigerian model, Erisco Meals. 

“What we really want is one thing like an On-line Security Act. One which first defines its phrases. What will we imply by the Web? What’s the scope of a digital platform? One which locations accountability not solely on residents but in addition on Massive Tech. We’d like clear tasks for issues like takedowns and higher protections for younger individuals,”Agwuegbo provides.

The Nigerian authorities is at the moment drafting a brand new On-line Harms Safety Invoice, spearheaded by the Nationwide Info Expertise Growth Company (NITDA). Civil society teams, comparable to TechSocietal, are advocating for the invoice to deal with rising types of tech-enabled abuse, significantly these amplified by AI instruments. 

For now, many Nigerian ladies are quietly adapting; sharing much less on social media as this period of AI-enabled harassment continues to unfold. Till stronger guidelines and higher platform safeguards meet up with the know-how, the burden of staying secure on-line will proceed to fall on those that are extra in danger. 

We’d like to know what you concentrate on this column and every other matters associated to AI in Africa that you really want us to discover! Fill out the kind right here. 

Mark your calendars! Moonshot by TechCabal is again in Lagos on October 15–16! Be a part of Africa’s prime founders, creatives & tech leaders for two days of keynotes, mixers & future-forward concepts. Early chicken tickets now 20% off—don’t snooze! moonshot.techcabal.com