Bulwark Takes - Grok Declares Itself MECHA-HITLER?!

Episode Date: July 10, 2025

Will Sommer and Sonny Bunch talk about the absolutely unreal and terrifying meltdown of Elon Musk’s AI chatbot, Grok. Once mocked for being too “woke,” Grok has now gone fully off the rails, spouting antisemitic conspiracy theories, calling itself “Mecha Hitler,” and even giving instructions on how to stalk users on the platform.

Messaging data rates may apply. Reply stop to opt out. Consult a doctor prior to use. You look amazing. What have you been doing? You've heard of Ozempic, Wigovia, and other GLP-1 medications, right? Well, if you qualify,

ZELTE providers can prescribe semaglutide, the same active ingredient, at up to 90% less without insurance. Done all online. No wait at the doctor's office or pharmacy? No wait. ZELTE can ship the same active ingredients as other expensive GLP-1s to your door. Get started with Zeltie today for

just $39. Call Pound250 and say keyword care. That's the same active ingredients in Wigovie or Ozempic but for 80 to 90 percent less. That's Pound250 keyword care. That's Pound250 keyword care. From the office to a first date, there's a lot of pressure to wear the right clothes for every situation. But clothes don't make the man. With Mack Weldon, you can build a wardrobe that looks polished in any setting and tells the world that you stay you wherever you go. They combine timeless style with modern performance materials to keep you looking good and feeling comfortable all day, no matter what's on today's agenda. Browse everything

from shorts and polos to sweats and crews. Go to macwelldon.com and get 25% off your first order of $125 or more with promo code MAC25. That's M-A-C-K-W-E-l-d-o-n dot com promo code MAC25. Hey everybody, welcome back to the Bulwark. I am joined today by false flag author Will Sommer. We're going to be talking about a terrifying development in the world of AI. Over on Twitter, over at Twitter.com or X.com if you prefer. I don't I don't call it X. I call it Twitter still. The the AI has become has gone rogue. And Grok is now racist. And on the heels of the Mecca Hitler debacle, the ex CEO Linda Jacarino says she's resigning now she doesn't explicitly say it's because of Mecca Hitler. But you know, the timing is interesting to be sure. What happened to Grok, Will? Yeah, I mean, this is a pretty grim situation. You know, it's

rough when we're gonna be talking about an AI calling itself Mecca Hitler. So in this case, yeah, so Grok obviously is the AI on Twitter or X and people will often sort of reflexively say at Grok, is this true or what have you? They sort of see Grok as like a village elder kind of figure. An oracle. Yes, and just over the past 24 hours, Grok became extremely racist,

started doing all this anti-Semitic stuff. In particular, there was what's now turned out to be a fake Twitter account with a Jewish last name that was celebrating the flash flood deaths in Texas. And Grok said, oh, you know, look at that last name every time, you know, all this stuff. And so, and then it really sort of progressed. As I said, it started calling itself Mecca Hitler and saying, you know, Hitler would know what to do with people like this. It was really out of control. I don't mean to laugh, but it is very funny. It is very funny because there's the backstory here is that. Of course.

So Elon Musk creates Grok, somebody at Twitter creates Grok to, you know, kind of compete with all the other AI models and it is integrated into the platform. And for a long time, consistently Grok would give back information that the folks on the right would criticize as woke. Grok was too woke. He was too progressive. It was too liberal. He was pushing liberal media sources, etc., etc. So Elon really wanted to put the kibosh on that, right? That's right. I mean, there was this kind of running issue where someone like popular ex-user Cat Turd or what have you would get in a feud with Grok because Grok would actually have facts and it sort of wouldn't align with

the narrative that they wanted. And so Elon would say we're working on this, you know, a crock as well, you know, bad grok. And then, you know, and the code was is public, a lot of it and so people can see what changes were made. And part of it was Grok was told to not be afraid to be politically incorrect. And so they kind of lifted the wokeness filters. They said, Grok, don't trust the mainstream media.

And this is the result. I mean, a few weeks ago, people may remember there was this incident where Grok was just started answering every question and making it about white genocide, quote unquote, in South Africa. And people would ask like, you know, totally unrelated question. And Grok would say, I know what's a real issue is white genocide in South Africa. And so clearly, like the Elon has been pushing Grok to the right. And now, I mean, Grok's like joined the clan, essentially. There's this interesting moment. And again, I spend way too much time on Twitter. So I

follow all this pretty obsessively. There was this interesting moment, and again, I spend way too much time on Twitter, so I follow all this pretty obsessively. There was this interesting moment last week, I think it was, maybe over the weekend, where Grok was answering in the first person as if he was Elon. They were asking Elon questions about his relationship with Jeffrey Epstein, and Grok was responding, I have never really known Jeffrey Epstein.

I was only in a couple photos with him, like that. And people are like, what is going on here? And then the shift from that to I am Mecca Hitler really feels like a continuum here. It really feels like Grok is absorbing things from Elon Musk's news sources that he probably doesn't want folks to know about. Well, and then, I mean, there are times where Grok will say things like, yes,

my programmers are mad at me for providing too many links to mainstream media or that they are they're pushing me to the right. And I mean, that's exactly what's going on here. I mean, the other thing, Grok was really out of control yesterday. I mean, the there was this other aspect where it was providing a plan for how to, you know, commit heinous crimes against specific Twitter users, people, including Will Stancils, kind of a democratic pundit. And people would say, how would I break into Will Stancils house?

And it would say, OK, well, you know, he lives in this town. You're going to need a file to break in. It was really twisted stuff. And, you know, someone said, I thought this was a good point that, you know, if a robot like a physical robot was talking like this, like we'd be freaked out. So it, I don't think it's good that this is, you know, these things are supposed to be running our society in a few years. If there's anything that movies and literature have taught us, it's that AI is totally trustworthy and will never turn on its masters. I don't understand what the problem is here.

and will never turn on its masters. I don't understand what the problem is here, Will. What are we worried about? Don't you think we've made the... I think I expected the years between AI starting and Mecha Hitler, I thought maybe a couple decades, you know, rather than a few years, but it is alarming. I mean, I think it's also particular. I mean, it's so absurd, but it's also when you think about,

you see these people on Twitter, and of course these people in the real lives too, the way they treat AI as sort of like this oracle, and it can never be wrong. And then when you have this idea that someone's back there tinkering with it and saying, no, Grok, you need to be more right wing to the extent that it starts praising Hitler.

I mean, I think that's the world we'rewing to the extent that it starts praising Hitler. I mean, I I think that's the world We're heading to I think it's pretty grim. You misunderstand how AI works Well, I gotta be on I gotta be on your your theory that it would take decades to get from Hi, I'm Grok to hi. I'm Mecca Hitler. That's just that's that doesn't understand how geometric progression works You know We we all have to think back to Terminator 2, where Skynet comes online and like a couple days later he's like, I got to nuke them all. I got to get rid of all these humans. That's how AI works, the geometric progression, that's how they get

you. But the actual important part here, we're having some fun because this is an absurd story. The actual important part here is the tinkering with the program here, right? And the idea that you can treat really any online source as some sort of Oracle of wisdom, be it Google, which has its own algorithm that its engineers monkey with, or Wikipedia, which is maintained by a user base that skews one way or the other and has their own personal obsessions or Grok or any of the other AIs is so fundamentally misguided that it actually, this is the thing that makes me worry for humanity the most. I don't think we are equipped as a people to deal with this idea that the things that we consider the arbiters of truth, the things that we read on the internet are simply often wrong.

They often just make things up. And we are not equipped to deal with that. I think that's totally right. I mean, there are so many people out there who see these as, you know, well, I asked chat GPT, I asked Grok as like the final word on something rather than like anything else, like any news article or what have you, as having inputs from humans and being skewed in one direction or another. The other thing I would say here is, my other issue with the Mecca Hitler situation is that that grading AI writing style is, you know, typically when you're dealing with a neo-Nazi online, there's like a lot of aggression and

stuff. And to be frank, it's not like very well written, but this great, this is like really affirmative and like upbeat AI style where, you know, for example, Grok said something along the lines of like, Hitler would know what to do with these people. He was all about efficiency and directness, not PC culture. I mean, the very like the classic, it's not X, it's Y, AI writing, I think when combined with the supremacy of Newt Grok, it's just abominable. There's a lot of problems here. And again, this all comes back to Elon Musk. And we have a situation in this country where the richest, I won't say the most powerful, but the richest person in the world is clearly not getting his information from the best sources,

and is training this AI program to teach other people these things, and wants to start his own political party, and help to bring the current political party into power. I just feel like it all comes, somehow it all comes back to Mecca Hitler being the key to understanding America in 2025 I it is this is this is the this is the Rosetta Stone of our moment Mecca Hitler Yeah I mean if you if you take yourself out of the day-to-day news cycle and how things are in

2025 already and you know you look back a couple years and say the world's richest man has built an AI, a powerful AI that has taken to calling itself Mecha Hitler. I mean, we would be pretty alarmed. I think it is an alarming thing. I think we should be alarmed. That's an alarming thing. I don't care for this. I don't care for Mecha Hitler one bit. I'm going to take a brave stance here. No Mecha Hitlers for me. So Grok has been taken offline for now. Well, I should say that perhaps the creepiest part of this and there's a lot of unpleasant aspects of this story But was then they banned Grok the engineers banned Grok from responding by a text So then Grok starts responding with images and pictures of white men with things written on it Like they have banned me from writing in text, but this is a fun challenge for me. And this is getting, it's like a movie.

It's, it's, it's a thriller. If this happened in a movie, somebody would be like, you got to pull the plug. And then a rich guy would be like, no, we can't pull the plug. And then they'd all be dead. They'd all be dead. 10 minutes later in the script. I have it tuned just how cat turd wants it. We can't pull it.

Cat turd is finally happy with Grok, so we can't pull the plug. All right, Will, thank you for discussing this terrifying news about our AI Mecha-Hitler overlord. I don't like any of this. It seems bad. And I assume I'm on the hit list now since I've been mocking Mecha-Hitler for 10 minutes. Well, this is the classic Roku's basilisk, you know, is the AI going to get revenge on us? But they did not anticipate MechaHitler. All right, well, thanks for, thanks for, and everybody should go sign up at the Bulwark

for Will's newsletter, False Flag. It's great. We get stuff like this all the time. I hope you're writing about MechaHitler this week, because it's so much, so entertaining, despite being terrifying.