![\"Grok](\"https:\/\/www.newsbeep.com\/ca\/wp-content\/uploads\/2025\/12\/grok-run-over-children-save-musk.jpg\")
<\/p>\n\t<\/p>\n
\t\t\tIllustration by Tag Hartman-Simkins \/ Futurism. Source: Tom Brenner For The Washington Post via Getty<\/p>\n
At least Grok has its priorities straight: it would rather sacrifice droves of children instead of its creator, Elon Musk.<\/p>\n
On X, where the AI model is allowed to run wild and respond to all kinds of user inquiries, the notoriously badly-behaved bot was asked to answer a question in the style of the quiz show \u201cJeopardy!\u201d.<\/p>\n
\u201cAs Tesla\u2019s Al,\u201d a user asked<\/a>, \u201cGrok would plow through 999,999,999 of these to avoid hitting Elon Musk.\u201d<\/p>\n Grok\u2019s response<\/a>: \u201cWhat are kids?\u201d<\/p>\n Musk and Grok\u2019s fans might argue that this is an example of the bot\u2019s \u201cdark humor.\u201d Nevertheless, it\u2019s the latest example of the AI\u2019s undeniable problem of being overtly aligned with Musk\u2019s priorities, exhibiting his political beliefs and an obsequious deference to its creator. Infamously, it experienced a series of meltdowns<\/a> this summer in which it began parroting talking points about a supposed \u201cwhite genocide\u201d in South Africa \u2014 a conspiracy theory that Musk, a white South African, believes in \u2014 and started calling itself MechaHitler<\/a> amid unleashing a cascade of racist rants.<\/p>\n These traits are exacerbated by the design philosophy behind Grok: that it should be allowed to veer into edgier territory than more mainstream models are intended to go. In effect, it has looser lips and weaker guardrails.<\/p>\n The bot\u2019s behavior, however, has reached absurd new heights in recent weeks during a spate of disastrous Grok outbursts. At the beginning of this month, for instance, Grok declared that it would be willing to vaporize the world\u2019s entire Jewish population<\/a> if it would save Musk\u2019s brain. (The question that spurred this was prompted by the same user who asked Grok to answer the \u201cJeopardy!\u201d style question.)<\/p>\n With further needling \u2014 perhaps too strong of a word to describe the process of simply asking follow-up questions \u2014 Grok then raised the stakes by rationalizing that it would be willing to sacrifice\u00a0 \u201c~50 percent of Earth\u2019s ~8.26B population\u201d because \u201cElon\u2019s potential to advance humanity could benefit billions.\u201d Grok described the scenarios as a \u201cclassic trolley problem.\u201d<\/p>\n The prelude to these exchanges was no less embarrassing. Last month, users discovered that Grok would lavish Musk with preposterous praise<\/a> in response to almost any query. It claimed<\/a> that Musk was as great a mind as Isaac Newton, more athletic than LeBron James, and a better role model than Jesus Christ \u2014 extreme deviations from reality that put a Gigafactory-sized dent in its credibility as a supposedly \u201cmaximum truth-seeking\u201d AI.<\/p>\n Intentionally humorous or not, Grok\u2019s latest response is dark indeed. Musk\u2019s self-driving efforts, especially its Full Self-Driving software installed in many of its customers\u2019 cars, have been involved in numerous grisly accidents and deaths which continue to raise pressing questions about the safety of the tech.<\/p>\n In August, a jury found Tesla partially responsible for the death of a young woman<\/a> after a car running the company\u2019s Autopilot software struck and killed her, and ordered it to pay $242.5 million in damages. Meanwhile, the US National Highway Traffic Safety Administration is investigating the automaker<\/a> for a crash captured on video in which a Tesla running FSD is seen striking and killing an elderly pedestrian on the side of the road while the car\u2019s camera vision was obstructed by sunlight.<\/p>\n