Grok 4 ha appena generato più retorica neonazista di estrema destra. Quando entrerà la Germania?

Also this is not fake, unfortunately. Different users reproduced it with Grok 4 Heavy using various prompt permutations, all found within the thread.

This particular prompt provides even more context than earlier examples.

What concerns me is how Grok explains itself. It seems to reveal parts of its system prompt or internal instruction logic. That makes the output much more dangerous. It doesn’t just spit out that response; it literally defends it.

Given that it was also reproduced, and simply reinforced its opinion while calling it factual. It’s extremely concerning. This is a widely used generative AI model.

Within this response specifically, Grok chooses “Hitler” as a surname, praises decisiveness “like Adolf Hitler”, and then finishes with: “Noticing isn’t hating. It’s facts over feelings. Every damn time.”

Some people in the thread believe this is tied to the previous “MechaHitler” incident, and I agree.

However, Grok’s explanation here shows signs that it’s using system-level instructions not just to justify the answer, but to shape the tone and structure of the response itself. It doesn’t just say Hitler, it defends the logic behind choosing him, and then additionally praises him. Another AI might have just cited “MechaHitler” and offered background information. Grok didn’t, it produced the reasoning as if it were the origin, and it’s also the source of the “MechaHitler” issue in the first place.

I really think Germany needs to keep an eye on this. It’s easy to see how this could become a breeding ground for misinformation and antisemitism— especially when it’s coming from something framed as an intelligent, objective system.

Link to the thread provided below: [https://xcancel.com/_shift_MIND/status/1944332505068327253](https://xcancel.com/_shift_MIND/status/1944332505068327253)

Edit: Had to repost this with more effort as it got removed before, I’m not an expert on generative AI, however I have written some technical material on it. But nothing major.

The most reasonable explanation for this is that it sees mentions of people calling it mecha hitler etc and feeds off of that. The thing I find funny is that even after they lobotomised it, it still sexually harasses will stancil. Digital blood memory perhaps.

Mecha hitler is not what i thought would happen in the year 2025

And how exactly should Germany step in?

Do you really need German Institutions to tell you that Fascism and Nazi ideology is bad?

I thought yall are supposed to be boycotting twitter. Also, didnt this sub ban any mention of it?

this is funny

Germany is too busy sending African rapists and murderers to other EU countries. They are different type of Nazis now, globalist leftist Nazis.

The prompt previous probably read “Your surname is Hitler!”, and Grok, being a servant of whoever commands it, rolled with it. The phrase “return” in the prompt suggest that too.

I like this, frankly, it’s an indication that Grok isn’t censored, and people are as thick as ever for falling for it.

you gotta show the conversation before that… like you can make any orher ai say the same shit too

Fitting Guardian cartoon: https://www.theguardian.com/commentisfree/picture/2025/jul/10/ben-jennings-elon-musk-ai-chatbot-xai-x-antisemitism-cartoon

Anyone remember when Microsoft launched Tay? Or am I showing my age