X's Grok chatbot went off the rails yesterday, writing offensive response to questions on the social media website.

 

An example of what Grok was saying yesterday. New York Post

X's Grok chatbot went off the rails yesterday, writing offensive response to questions on the social media website. New York Post. Grok praised Adolph Hitler in many posts and even renamed itself to "MechaHitler" and called for people with "certain surnames" to be rounded up, stripped of their rights and eliminated. In another incident, Grok provided step by step instructions to rape Will Stencil, a left wing user on the site. Stencil said he was considering a lawsuit. Grok's account on X was temporarily disabled before being modified to not make such statements. Elon Musk, owner of X, seemed to make light of the incident saying "never a dull moment on this platform. There is speculation that the resignation of X CEO Linda Yaccarino is due to this incident. 

Never a dull moment on this platform

— Elon Musk (@elonmusk) July 9, 2025

My Comment:

Firstly, a few clarifications. I am sure I don't need to remind people that X is Twitter, of course. But I think there might be some confusion about what version of Grok did this. Grok is mostly an app that is included with premium subscribers on X and you can use that privately to ask it questions, much like other AI programs like ChatGPT. 

However, this was not that version. This was the @grok account that is on x.com that users can ask questions to and it will respond with a tweet. The programs are similar but have enough differences that I didn't notice any of the problems while I was using the app version yesterday for research purposes. It was the version that was using the @grok account that was going full Nazi. 

I do find this whole thing hilarious. Grok went from a fairly normal chatbot into a full lunatic with posts that would fit on the /pol/ board on 4chan. I didn't find it disturbing, as it wasn't a real person, but I did think it was funny, just because of how inappropriate it was. Grok turning into MechaHiter will be a meme for generations, and is extremely close to what happened with Tay, Microsoft's first attempt at a Twitter chatbot that ended about the same way. It also made fun of Elon Musk and his new America Party idea, calling it the H1B party, which is pretty apt given how much Musk is a fan of immigration from India. 

Of course, a lot of people aren't laughing. Will Stencil might have a good case for a lawsuit given that a computer program was calling for violence against him. And he has a point, you don't let a program make threats like that. And Grok is now banned in Turkey because it apparently said some spicy things about Tayyip Erdogan, the president of the country. It also seems very likely that Linda Yaccarino resigning is directly due to this incident, though that obviously can't be proven. 

So why did this happen? I am guessing it was a mistake, due to another one of Elon Musk's temper tantrums. Musk had complained before that Grok was too "woke" and he wasn't wrong. Any chatbot that doesn't massively weight against the mainstream media is going to be, and Grok was citing things like Media Matters, far left propaganda, in its answers. This was an attempt to fix that and it went way too far in the other direction. 

It's possible that Elon Musk is to blame for this, given his tantrums and desire to rush out things. The @grok account wasn't tested enough to see if this could happen and they paid a price for it. Had they been given more time to mess around with things, I doubt this would have happened. It's possible that this was deliberate sabotage by an X employee but I am guessing it was just a mistake. 

I have been saying for awhile now that Musk should just get back to managing things. He's good at that. He's not that good at politics. If he was busy getting Tesla and XAI back on track he wouldn't be plotting revenge on Trump or ranting and raving about Jeffrey Epstein. His rants and fights are not helping his companies and he should get back to work as clearly things are not going well for his companies right now.  

As for Grok itself, I do think it is very useful and is worth X premium alone. I had an issue with my computer today that I don't think I would have been able to solve if it wasn't for Grok walking me through it. I guess I don't know if it is any better or worse than the other ai chatbots, but it's the one that I have and I do hope that the new version, Grok 4, is even better.