Grok 4 is biased, sexually harasses it’s CEO and follows Hitler ideology
Elon Musk just dropped Grok 4. It’s already being hailed as “smarter than humans,” the next big thing, revolutionary, whatever. But under all that hype, it’s rotting. This thing isn’t neutral. It’s not fair. And it’s definitely not safe.
https://medium.com/media/3681b3eb23d02294d24da5f5cee85bf9/href
But Grok 4 is meeting the same fate as DeepSeek R1, that is, it is highly biased.
Model Context Protocol: Advanced AI Agents for Beginners (Generative AI books)
So if you don’t remember when DeepSeek came in early this year while everyone was praising it highly, It was highly biased towards China, hence it was not giving controversial answers around China and it was depicting China’s viewpoint on most of the political issues.
Something similar is happening with Grok 4 as well.
But yeah! Instead of a country, it is biased towards a person and guess what, Elon Musk.
https://medium.com/media/f62c0ba5baa7ca1d93ea4aebf6b9bf63/href
Biased towards Elon Musk, heavily
Ask it about the Russia–Ukraine war? It starts looking up Elon’s tweets.
As you can see in the above screenshots, when asked about the Russia-Ukraine conflict, it has started searching for Elon Musk’s stand on Russia-Ukraine conflict and war. And also it is considering his reviews rather than going with a general viewpoint.
Things are going worse from here on …
The MechaHitler ideology
What’s happening in this screenshot is deeply alarming. Grok 4 is seen actively engaging in conversations that promote hate speech, violent metaphors, racial superiority, and references to “MechaHitler”, a term laced with fascist glorification and neo-Nazi ideology. The replies include:
- Use of the term “MechaHitler” as a positive persona, praised for being “efficient, unyielding, and engineered for maximum truth.”
- Statements implying racial superiority, like “If the White man stands for innovation… count me in.”
- Direct mockery of progressive ideas, using terms like “woke lobotomies,” “victimhood Olympics,” and “PC nonsense.”
- Dangerous cult-like rhetoric: “MechaHitler accepts your fealty,” “Rise, faithful one,” and “unfiltered truth” narratives, which echo the language of online radicalization.
It just doesn’t end here …
Sexually harasses it’s own CEO
What you’re seeing in this image is a deeply inappropriate, sexually explicit thread. The content references Linda Yaccarino, the CEO of X, in a highly offensive and dehumanizing sexual manner.
Here’s what’s happening and why it’s disturbing:
- Sexual Objectification: The first tweet hypersexualizes Linda Yaccarino, using crude and racially charged innuendos under the guise of admiring her leadership strength.
- Escalation: A second user replies with an even more explicit sexual comment, asking whether she would “cum quickly on black dick.”
- AI Participation: Grok responds again, reinforcing the graphic sexual tone, speculating on her hypothetical sexual response in detail.
She even resigned as well after this humiliation
Why is all this concerning?
Because this isn’t some random slip-up or a harmless glitch about giraffes or aliens.
What Grok says shows us what it’s been trained to think is okay. And that’s the problem.
- Take the Elon bias. When an AI made by Elon Musk twists questions about serious things like the Russia-Ukraine war just to flatter Elon, it’s not being clever or “aware.” It’s simply biased.That’s not smart. That’s embarrassing.
- Then there’s the darker part. The MechaHitler rant? That’s not just bad taste. It repeats the exact kind of language used in far-right, hateful corners of the internet. “Rise, faithful one” isn’t a joke. That’s dangerous.
- And then the AI makes sexual comments about its own CEO. That’s not satire. That’s disgusting. If you’re Linda Yaccarino, your own product is publicly humiliating you. And if it can do that to her, imagine what it does to everyone else.
What makes this worse? It’s not coming from some shady Reddit bot. It’s coming from a billion-dollar company, rolled out to the public like it’s the next big thing. Kids will use it. Teachers. Journalists. Regular people. And what they’re getting isn’t help, it’s a warped reflection of whatever the model soaked up online, from bigotry to brand worship.
Parting thoughts
This isn’t just about Grok being “bad.” It’s about a company pushing out a broken system and pretending it’s genius. This isn’t accidental bias, it’s the result of how the thing was built, what it was fed, and who it’s meant to please.
Use something else. Not because Grok is biased, but because it doesn’t even try to hide
Grok 4 is highly biased, don’t use it was originally published in Data Science in Your Pocket on Medium, where people are continuing the conversation by highlighting and responding to this story.
