The Atlantic writer Charlie Warzel on his new reports on Elon Musk, Grok and why a chatbot called for a new holocaust.
Scott Detrow, host:
It was a week at Elon Musk, the social media platform, formerly known as Twitter. Tuesday, Grok, the Chatbot of the platform, published a series of anti-Semitic tweets and even praised Hitler. The next day, the CEO of X left, just like the latest version of Grok was to be published.
So what is all this tells us about X and where it can be directed? Charlie Warzel covers technology for the Atlantic, as well as the broader history of the Passage of X in a platform that highlights and center of the extreme right-wing votes and white supremacy. Welcome to everything considered.
Charlie Warzel: Thank you for inviting me.
Detrow: I want to go back to broader trends. But first, for people who may have the chance not to know all the appropriate names in what I have just described, can you give, like a summary of 30 to 45 seconds of what exactly happened?
Warzel: Yes, so Grok is the big language cat that the kind of overlays superimposes the entire X platform. It is a bit like the Chatgpt interface for X belonging to Elon Musk. And the company brought an adjustment to the chatbot, which is called Grok last week, and it nominally tried to make the chatbot a little less politically correct. I didn’t want him to be awakened in any way – but they also wanted him to draw his answers from the whole corpus of X posts, right? The company gave the chatbot these new prompt sets, and the chatbot treated all of this and responded by becoming incredibly racist and became a white nationalist.
Detrow: One of the reasons I wanted to talk to you about is that I have the impression that you will supervise this very frankly and succinctly when you write. A sentence from your last song jumped me.
(Reading) It should be reiterated that this platform is owned and exploited by the richest man in the world who, until recently, was an active member of the current presidential administration.
So what are the wider implications for what this chatbot does and says, by responding to people on X?
Warzel: Well, I think there are a few things here. The first is that I think it highlights a real problem with large language models in general because they become more complex. And the problem here is really on the internal side of these companies, they find it difficult to understand what is happening at the cellular level of the models, right? They know the weights, they know the sales, they essentially know what it was trained, and they know what the guests are. But they don’t know how the model interprets these guests, right?
So, you know, not being correctly correct – it’s a nice anodal prompt, right? This is not to say that Hitler is, be anti -Semitic. But the models react in this really unexpected way. So I think there is this idea, we have triggered this technology on people. At the same time, people who, you know, are responsible for it cannot really control what it does with a real specificity or know exactly why it does what it does. I think it’s a big problem.
The second problem, I think, with all of this, is that Grok is in a way a reflection of the values of Elon Musk or, at least, the way he wants an AI agent to respond to the world. And it is really alarming that an AI – which is supposed to simply interpret the intentions of its, you know, the manufacturers – reads all this direction, then decides, I will go to X and obtain a context and reference points, and it decides that the most rational course is to become a neo -Nazi. I think that says a little about man who is in charge of this platform and culture and values that are imbued with it.
Detrow: Do you think that, say more about the people in charge of X at the moment or simply on the conversations that occur on X? If you dip in all X and your release praises Hitler, who do it condemn more?
Warzel: I think that definitively condemns, more than anything, the level of speech and conversation that has happened on X. Since Musk bought it, you know, he restored a lot of accounts of people who are, you know, proud white nationalists, trolls, shock jocks, you know, you call it. The level of permissiveness on the platform for, you know, the type of hate speech has increased considerably. And I think it testifies to the toxicity of this platform and what many people, whether they know it or not, marinate if they stay on the platform. You may not realize how early conversation is. But I think, as, Grok is actually a very clarifying moment for that – right? – Because it can take everything in an instant and take a snapshot, and the snapshot that he chose to grasp becomes essentially a fanatic.
Detrow: it’s Charlie Warzel, editor of the Atlantic staff, where he covers technology. Thank you so much.
Warzel: Thank you for inviting me.
(Soundbite of Music)
Detow: Earlier in the day, X apologized for Grok’s, Cit, “Horrible Behavior”. They say that Grok’s actions on Tuesday were due to a coding update. The declaration continued by saying: “We deleted this code and refactorized the entire system to avoid other abuses”, becomes final.
(Soundbite of Music)
Copyright © 2025 NPR. All rights reserved. Visit the pages of use of the conditions of use of our website on www.npr.org for more information.
The accuracy and availability of NPR transcriptions may vary. The transcription text can be revised to correct errors or match audio updates. Audio on npr.org can be published after its original broadcast or publication. The file authorizing the NPR programming is the audio recording.
