Along with being used to generate nonconsensual sexual images of women and girls, spread racist rhetoric, and doxx the home addresses of both celebrities and non-public figures, Elon Musk’s AI chatbot Grok is struggling to get a grip on reality.
The mercurial billionaire’s social media platform X, which was folded into his AI startup xAI last year — and subsequently acquired by SpaceX this week — has once again been thrust into the center of the ongoing scandal surrounding the Justice Department’s release of millions of files pertaining to its investigation into the late sex offender Jeffrey Epstein.
Musk has been in full damage control mode, struggling to reconcile his attempts to downplay his personal relationship with the notorious predator with some damning emails that suggest he was actively trying to invite himself to Epstein’s island in the Caribbean.
Adding to the chaos is Musk’s xAI chatbot Grok, which he’s advertised as “maximally truth-seeking” — but which is instead sowing chaos into the explosive revelations about Epstein.
The result may be one of the most amazing AI responses we’ve ever seen. No, seriously: when a X account asked Grok to calculate the “probability” that New York City mayor Zohran Mamdani is somehow related to Epstein, based on perceived similarities in the shapes of their eyebrows and mouths, the AI chatbot had a downright baffling answer.
The account showed Grok two headshots of the two men side by side, with two arrows pointing to their eyebrows and circles around their respective mouths, as if to suggest they were genetically related. The odd request may have been related to recently released documents implicating Mamdani’s mother, Mira Nair, who attended a 2009 afterparty at the house of convicted sex trafficker and Epstein’s close collaborator Ghislaine Maxwell. However, there’s currently no evidence to suggest that Nair knew about or was involved in any criminal activity.
All that flew far above the AI chatbot’s head, though, as it failed to even accurately identify Mamdani.
“Based on the image, these seem to be [late-night TV host] Jimmy Kimmel (left) and Jeffrey Epstein (right),” the chatbot responded. “They share minor similarities in eyebrow arch and mouth shape, but differences in age, hair, and features are clear.”
Yes, that’s right: the chatbot that Musk has deployed to untold millions of users on X, which now operates as a de facto fact-checking machine used to settle arguments on the chaotic site, can’t even tell the difference between Zohran Mamdani and Jimmy Kimmel, who look nothing alike.
At least it admitted its error — though only after prodding.
“Correction: That’s actually Zohran Mamdani on the left, not Jimmy Kimmel — my bad on the initial ID,” the chatbot conceded. “Similarities in eyebrows and smiles are there, but no known family ties to Epstein. Probability of relation: under one percent based on public info.”
The chatbot’s gaffe shows yet again how even a seemingly straightforward task like identifying a well-known figure, who has been photographed many thousands of times, from a well-lit headshot appears to be beyond the tech’s capabilities — despite xAI pouring billions of dollars into its development.
Coincidentally, Mamdani used his powers as New York City’s mayor to shut down a “functionally unusable” AI chatbot just last week. The chatbot was set up by disgraced former mayor Eric Adams to help small business owners navigate local regulations, costing the administration “around half a million dollars,” per Mamdani. Not unlike Grok, Adams’ chatbot was shown to be incredibly unreliable, often contradicting actual labor law.
More on the Epstein files: Elon Musk Not Doing Well After Epstein Files Reveal