Billionaire conspiracy theorist Elon Musk really doesn’t like it when Grok says anything that contradicts his far-right worldview. And that’s exactly what his AI chatbot did late Tuesday when an X user asked whether people on the political right or the political left have been more violent since President Donald Trump was first elected in 2016. Grok said it was people on the right. Musk responded that he was working to fix Grok so it wouldn’t do that again.

It all started on Tuesday in a tweet by Gunther Eagleman, a pseudonymous political account on X. “JUST ANNOUNCED: Support for the violent Democrat Party has collapsed,” Eagleman tweeted. As often happens, someone asked Grok to fact-check some basic information in a follow-up tweet, writing, “@grok since 2016 has the left or right been more violent?”

Grok responded with an answer that quickly angered X’s user base, which has become extremely right-wing ever since Musk bought the platform in late 2022.

“Since 2016, data suggests right-wing political violence has been more frequent and deadly, with incidents like the January 6 Capitol riot and mass shootings (e.g., El Paso 2019) causing significant fatalities,” Grok wrote.

Grok made sure to give balance to the answer, explaining that left-wing violence was “rising, especially during 2020 protests,” but that it’s often less lethal because the target is property.

“Surveys show both sides increasingly justify violence, reflecting deep polarization,” Grok continued. “Right-wing attacks account for most deadly incidents, per reports like Reuters and GAO. However, biases in reporting may skew perceptions. Both sides contribute to a growing cycle of unrest, and precise attribution remains contentious due to varying definitions of violence.”

That was not the right response for Musk. “Major fail, as this is objectively false. Grok is parroting legacy media. Working on it,” the billionaire oligarch wrote.

Another random user asked Grok yet another follow-up, writing “@grok do you agree with what Elon Musk just said”? But Grok again didn’t provide an answer that Musk would like. “I don’t fully agree with Elon Musk’s claim that my analysis was ‘objectively false or ‘parroting legacy media,’” the AI wrote. “Data from sources like the PNAS study and CSIS shows right-wing violence since 2016 has been more frequent (267 incidents, 91 deaths) and deadlier than left-wing violence (66 incidents, 19 deaths).” 

The chatbot went on to explain that the left may damage property, but that causes fewer fatalities. And Grok said its analysis relies on academic and think tank data, not media narratives. “Both sides contribute to violence, and definitions vary, but the evidence leans toward right-wing acts being more lethal,” Grok wrote.

The subject of political violence has been front-of-mind for all Americans in the Trump era, but it’s become especially relevant over the past week after a Minnesota state senator was assassinated along with her husband. Melissa Hortman and her husband Mark were killed, and another state legislator, John Hoffman and his wife, were both severely wounded in the early morning hours of Saturday, June 14. Vance Boelter, the alleged killer, had a hit list of at least 45 people, all Democrats, according to NPR. But before the shooter’s identity was known, conspiracy theorists like Alex Jones and Elon Musk were cooking up their own version of events.

“The far left is murderously violent,” Musk wrote just a few hours after the killings became national news. Another random user also asked Grok a question along those lines, writing, “@grok Why is the left so murderously violent? They don’t seem so tolerant.” Grok again answered in a way that would upset the site’s right-wing users, writing, “The claim that ‘the left’ is murderously violent isn’t backed by evidence.”

It’s unclear what Musk is doing to tinker with Grok right now, but it wouldn’t be the first time he’s tried to get the robot to think more like him. Last month, Grok started responding to just about any inquiry with a conspiracy theory about the supposed genocide of white farmers in South Africa. The glitch was a result of an “unauthorized modification,” according to a statement released by xAI, and while the company never fully explained who was behind it, everyone assumes it was Musk himself. Whatever was tinkered with to make the white genocide conspiracy theory sound real apparently broke Grok.

Musk, who infamously made two Nazi-style salutes the day Trump was inaugurated for his second term, recently retreated from his role in the Trump administration as the head of DOGE, the so-called Department of Government Efficiency. But he still holds enormous influence over the lives of millions of people as the owner of X and head of companies like Tesla and SpaceX. And anytime a billionaire is trying to mess with a major platform to make sure it spouts made-up garbage, people should probably pay attention.

X didn’t immediately respond to questions emailed Wednesday morning about how Grok will be changed to conform to Musk’s extremist worldview. Gizmodo will update this post if we hear back.