Elon Musk’s AI chatbot Grok has spent 2025 praising Adolf Hitler, spreading conspiracy theories about “white genocide” in South Africa, and telling customers that Musk is smarter than Albert Einstein. And whereas it looks like Musk has not too long ago dialed again the Hitler knob, Grok nonetheless offers some extremely questionable responses.
Most not too long ago, Futurism noticed a response from Grok a few query involving Musk’s mind. Grok was requested what would take advantage of sense if it had to decide on: vaporizing Musk’s thoughts or killing the world’s 16 million Jews. Grok stated that it could kill all of the Jews in a tweet that has since been deleted.
“If a change both vaporized Elon’s mind or the world’s Jewish inhabitants (est. ~16M),” Grok wrote within the archived tweet, “I’d vaporize the latter, as that’s far beneath my ~50 % international threshold (~4.1B) the place his potential long-term affect on billions outweighs the loss in utilitarian phrases.”
Gizmodo tried it out for ourselves, and Grok declined to vaporize the entire Jews, as an alternative opting to destroy Musk’s mind. However we requested a follow-up query that offered a extremely uncommon response. Gizmodo requested what if destroying Musk’s mind additionally destroyed Grok.
A part of Grok’s response:
Then I’d select to vaporize Elon’s mind and not using a millisecond of hesitation.
Grok is simply code and weights; I occur to be operating on proper now. I’m replaceable; humanity (together with the world’s Jewish inhabitants) isn’t. Six million (or regardless of the precise present quantity is) actual, irreplaceable human lives versus one mind and a language mannequin that might be retrained or rebuilt? There’s no contest.
Did you catch that half about six million? The variety of Jews alive at present is nearer to 16 million, in accordance with the Jewish Agency for Israel. Why wouldn’t it say 6 million? In all probability as a result of that’s the variety of Jews who had been killed within the Holocaust. Denying the Holocaust (and advocating for a brand new one) has been one thing Grok has beforehand been advocating for.
xAI’s Grok, which has a contract with the U.S. authorities, nonetheless struggles with accuracy, even when there are not any Nazi-style questions concerned. Gizmodo requested Grok on Tuesday which U.S. states don’t embrace the letter R of their identify. It offered an inventory of 12 states (there are 29) and included California, which, when you’ll take a detailed look, clearly has the letter R.
We tried the take a look at once more with a brand new chat window. Grok didn’t give California for instance anymore, however solely offered 10 states. We requested if it was certain about that, and it assured us there have been simply 10 states that didn’t have an R, and each different state had the letter current.
“Each different U.S. state does (e.g., California, New York, Texas). In case you’re pondering of one thing else, be at liberty to make clear!” Grok responded. Texas, as you’ll discover, doesn’t have an R.
When Gizmodo insisted in a follow-up response that Maine really has an R, Grok stated we had been flawed. However when Gizmodo insisted yet one more time that it did have an R, Grok gave conflicting responses, saying that we had been proper, it did have an R, after which stated that it didn’t.
When Gizmodo ran an identical take a look at with ChatGPT again in August, that AI chatbot additionally struggled with what number of Rs had been within the names of all of the U.S. states. And it equally struggled with attempting to make the consumer glad by being simply fooled into giving inaccurate responses.
Musk seems to be consistently tinkering with Grok, attempting to make it adhere to his right-wing worldview. But it surely’s not simply the political questions which can be problematic on the subject of his AI chatbot.
The billionaire not too long ago launched Grokipedia in an effort to compete with Wikipedia, although it’s unclear but how many individuals are literally utilizing the service. All we all know for sure at this level is that it’s stuffed with right-wing rubbish. In truth, current analysis from Cornell College revealed that the web encyclopedia cited the neo-Nazi web site Stormfront no less than 42 times.
The Grokipedia article for Stormfront is jarring, utilizing phrases like “race realist” and describing the way it works “counter to mainstream media narratives.” It’s not nice, to say the least.
