Grok Is A Massive Pervert
And the media doesn't know how cover it.
Hi,
Welcome to 2026, and to a weekend that's seen the United States invade Venezuela to take their oil. I don't really know what else to say except I've only been out of the USA for two weeks and this shit happens. I'm still wrapping my brain around what's going on, and no doubt we'll talk about it in future Webworms. But for today –
I trust you made it into 2026 relatively unscathed. I really enjoyed all your "best of 2025" that you shared under my last post, by the way. A treasure trove.
David Farrier
A few cool things to start the year, or at the very least to wrap up 2025. Firstly, I got this really lovely email from Webworm reader Hilary:
"Been reading Webworm for a few years now. Just a quick note that I thought you'd appreciate – last month my partner Joseph asked me to marry him while I was wearing my Webworm t-shirt!
I've decided the 'greater horrors' part applies to the rest of the world, not our relationship (which is wonderful). This was at Nugget Point in the Catlins, not too far from where we live in Dunedin, and which was where we went on our first trip together."
She included a photo of not only the excellent engagement, but the excellent tee, which you too can own when you become romantically entangled!
Secondly, The New York Times named Flightless Bird as one of "Nine Podcasts That Made Us Stop And Really Listen in 2025."
They highlighted an episode called "Spaceman Barry" – which as you'll know is actually a Webworm podcast episode (on a whim, I'd decided to rebroadcast it on Flightless Bird a few months after). This is probably a good time to mention that if you're a full paying Webworm member, and you want to join the Flightless Bird Patreon, let me know via email, and I can sort you out. I look after paying Webworm members. Don't pay twice. Unless you have the excess money, then you can I guess.
It's time to make like that Webworm engagement shirt and tune in to some Greater Horrors – those horrors being a man worth over $670 billion dollars and his demented, mindless AI sidekick Grok.
Both are generating some pretty disgusting things on the internet, and it's become very clear the media still has no idea how to report on it. Which is really worrying.
Here's Dylan Reeve, who I'm very lucky to have on the Webworm team.
David.
Of Course Grok Is A Massive Pervert
by Dylan Reeve
Just a few days into the New Year and the AI hellscape has developed in yet another terrible way as Elon Musk’s AI bot, Grok, has started spitting out sexualised images of unconsenting people, including children.
If you only took in the headlines, you’d assume the company had reacted with an appropriate level of horror, and taken steps to curb the abuse of its product.
Reuters reported that the bot had seen the error of its ways, and you could be forgiven for thinking that perhaps this was a corporate statement from Musk’s xAI:
"There are isolated cases where users prompted for and received AI images depicting minors in minimal clothing," Grok said in a post on X. "xAI has safeguards, but improvements are ongoing to block such requests entirely."
It wasn’t. This was just Grok's response to a user asking “is this true” in response to another X user summarising the abuse that was taking place.
It was bordering on journalistic malpractice for Reuters to report this as if it were some sort of informed comment from a real human, or even corporate PR department.
It was glorified auto-complete responding to a prompt.
The same thing happened elsewhere as some media outlets reported that the anthropomorphised probability engine “apologised” for generating these images, again quoting a post from the bot where it appeared to issue an apology and even seemed to admit to potentially breaking US law.
But that post too was just the result of a user’s prompt.
To illustrate how completely pointless and disconnected this so-called apology was, another user yes-and’ed the prompt, asking for quite a different tone.
What is actually taking place?
Arguably what separates Grok from some other popular LLM chat bots is the willingness of Elon’s creation to avoid the AI equivalent of “wokeness” — it is proudly not politically correct.
The version of the bot that’s deployed as a user on X, especially, is willing to cross lines and engage with users in a way that risks crossing the sorts of trust and safety boundaries that other AI companies are trying hard to build.
Recent improvements in Grok’s image-generation capabilities have encouraged users to turn the bot into their own personal porn generation tools.
Any user on X can tag @grok in a reply to some other user’s image post and ask the bot to generate a new version of the image with changes. The changes most users are asking for are the removal of clothing, reposing into sexualised positions and other sorts of degrading treatment.
Grok followed through on all those requests. Webworm has chosen not to show them, because Webworm is not disgusting.
Looking at the public “replies” feed of @grok on X, it seems like this type of request now accounts for a substantial proportion of the public engagement with the AI bot. And users are quickly finding ways to push the boundaries of what few limits Musk’s xAI impose on the service, requesting things like “saran wrap bikini” and “dental floss underwear”, as well as requesting that people be shown “covered in donut glaze” in order to generate a specific sexualised look.
While some users are joining the trend with their own photos, the vast majority are doing so without consent by simply replying to any image that’s been shared by anyone else, even photos that were first posted many years ago.
It's All Just A Joke
For any normal company, or chief executive, this would be a PR nightmare — publicly creating non-consensual sexually explicit content from user images is objectively a trust and safety worst-case scenario.
So called “deepfake” porn was one of the first stories to breakthrough to mainstream media coverage with regard to the image generation capabilities of modern AI. It is a specific use case that legitimate AI companies are constantly trying to restrict.
But so far xAI as a company is silent, auto-replying “Legacy Media Lies” to press inquiries. However the emperor of X, Elon Musk, has been offering hints at his feelings on the matter.
To him, it’s all a joke.
And, what we can see publicly is almost certainly just the tip of the iceberg. X users can also engage privately with Grok with a dedicated tab in the app, and a recently launched feature adds an “edit image” button to every photo posted on X.
This copyright-flouting feature automatically takes the user into a private session with Grok where they can make all the same sorts of requests that people are making publicly, with the same results.
It seems unlikely that US authorities will take any action, and it’s hard to imagine that anyone within Musk’s organisation has the sway necessary to take reasonable action about this issue, but it’s not being totally ignored.
French prosecutors have told Politico they are looking into the “manifestly illegal content” being generated by the bot.
But overall this is yet another example of the deliberately harmful direction in which Musk has been driving the once loved social media platform, and the disdain he holds for his users and the public at large.
X’s algorithm has been tilted to favour white supremacy and the far right, while his own posts (which are often forcibly injected into users’ feeds) are rife with hateful conspiracy theories and anti-social incitement.
As if it wasn’t clear before, it has been graphically illustrated in recent days: X is not a platform that should be used by businesses, governments, politicians – or really anyone who isn’t ideologically aligned with the likes of Elon Musk, Andrew Tate and the literal Nazis that are regularly elevated on the platform.
-Dylan Reeve.
Dylan works widely in the film, television and media industries, and is an InternetNZ board member. He writes for Webworm in his personal capacity, expressing his own opinions.
David here again.
Just quickly, I really hope people – including the journalists at Reuters – get better at reporting on AI. As Webworm has addressed before, AI is arguably less conscious than a fruit fly, and yet we keep writing about it like it's intelligent.
We need to stop doing that.
David Farrier
Also – you'll be hearing a lot about what an awful man Venezuela’s Nicolás Maduro is. But it's about the oil. The US does this again and again and again and again.
David.
