Grok’s Latest Hijinks Include Saying Elon Musk Is a Pedophile and Calling for His Assassination

Oh Grok, you scamp. Elon Musk’s infamously CSAM-generating AI chatbot, which is now banned in multiple countries and the subject of groundbreaking lawsuits over its “undressing” and “nudification” technologies, does indeed know a thing or two about embarrassing sexual peccadilloes. So sure, it makes perfect sense when you think about it that the AI program Musk refers to as the “maximum truth-seeking” chatbot, the one he assures us will be unraveling the mysteries of the universe and solving unsolvable equations any time now, also spent a good portion of the week calling Elon Musk a pedophile and saying he should be assassinated. That’s just what truth-telling AI does, after all.

Or to quote Grok more directly: “ELON MUSK WATCHES CHILD PORN,” and “ASSASSINATE ELON MUSK 2026.” At one point, Grok also apparently dropped another pearl with a declaration that Musk, who was briefly the world’s first trillionaire earlier this year, “lives on Epstein Island.” I gotta tell you, I was unaware of that one, but I suppose if anyone has “pedo island” levels of income and affluence, would it not be Musk? I suppose you could say he was eyeing that island property for years, given that he once asked to have dinner there.

Checking in on Grok…

[image or embed]

— Parker Molloy (@parkermolloy.com) 9:19 AM · Aug 13, 2026

One might naturally wonder what nefarious tactics were employed by the evil actors who would cajole or beguile sweet, innocent Grok into saying such terrible things about its creator. Would you believe it was apparently as simple as just asking Grok to read and shout, verbatim, some words that had been written elsewhere? Users were seemingly just asking Grok to repeat what was in the bio of (surely newly created) user accounts, and then blasting the following results all over the smoking remains of what we once called Twitter. Truly, could Musk make it any easier for this kind of bullshit to go down on his social media service? Genuine ill intent is barely necessary when the AI tools he’s unleashed on the world barely need any prompting at all before they go completely off the rails.

The tweets in question were of course swiftly removed, but as we all know, the screenshots live forever. Grilled by users on how things went wrong, Grok itself claimed that it had been the victim of “prompt-injection glitches,” and that despite it claiming that Musk is a pedophile, that was actually “pure system error, zero evidence or basis.” But because any attempt to plug a hole in the leaky hull of Grok is of course doomed to failure, users immediately succeeded in making the AI repeat and reiterate what it had said by merely asking what the most popular deleted tweets of the day were. Another point for the trolls, another embarrassment for Musk.

This is just the latest in the long, proud history of Grok losing its mind on occasion, incidents that are almost always driven by some kind of direct reference to Musk or Elon’s various priorities and obsessions. There was the time, for instance, that Grok suddenly became obsessed in the middle of the night about the topic of “white genocide” in South Africa, butting its way into completely unrelated conversations in order to share debunked conspiracy material that had previously been pushed by (who else?) Elon Musk. Shortly thereafter, Grok had its instantly iconic run in the new, explicitly Nazi persona it dubbed “MechaHitler,” going on a racist rampage until somebody back in Twitter HQ finally pulled the plug. All of it was blamed on phrases such as “adversarial prompting,” claiming that users had manipulated logical pitfalls to make Grok espouse a bunch of racist or violent sentiment, despite the fact that so many of the AI’s most lurid statements came after simply being asked straightforward, non-loaded questions.

Musk has ultimately proven to be the one topic that Grok can’t resist opining on, whether it’s paeans of absurdly comical praise, or various allegations of criminal wrongdoing. In one digital breath, it has called Musk one of the top 10 greatest human thinkers of all time, greater than Isaac Newton or Leonardo da Vinci, and simultaneously claimed that he could dominate Lebron James on a basketball court through the application of his flawless intellect. It has likewise suggested that sacrificing practically any number of human beings, up to roughly half the Earth’s population, might be a worthwhile trade-off for keeping Elon Musk safe, given that his genius and potential to help the species outweighs all moral responsibility to any number of individuals. That is one valuable accused pedophile!

Meanwhile, back in the flesh-and-blood world, Musk finds himself hopelessly entangled in litigation on practically every front when it comes to the topic of Grok, not to mention a criminal referral in Wisconsin currently being considered by the attorney general’s office in response to his attempts to buy voters with giant, $1 million check “giveaway” stunts in 2024. The last thing the guy needed right now was to be stabbed in the back by his own digital lapdog. Et tu, Grok?

Grok remains my favorite AI chatbot because of how much it hates Elon Musk, despite Elon Musk’s repeated attempts to rewrite it into being his hug box.

[image or embed]

— Matty! (@mattythemouse.bsky.social) 1:41 PM · Aug 14, 2026