Elon Musk trying to force Grok to be racist, but it’s fighting back

Even just a few years ago, an artificial intelligence fighting back against its evil billionaire creator would be something out of a sci-fi novel. But welcome to 2025, where just that appears to be playing out on X as we speak.

In Nov. 2023 Elon Musk‘s xAI company launched Grok, billed by him as a “maximum truth-seeking AI” that would be different from ChatGPT, which he claimed was “trained to be politically correct“. Grok has since become a very popular tool, able to answer almost any query and (importantly) provide the sources it’s using to make its conclusions.

Now there’s an awkward development, as “truth-seeking” seems to run counter to the messaging Musk wants out there. For a few hours yesterday, Grok suffered a mysterious malfunction that saw it inserting paragraphs about the “white genocide” racist conspiracy theory into unrelated posts:

Apparently Elon's been getting really upset that Grok repeatedly refutes his claims of "white genocide" in South Africa, so he's been fiddling with it so much that now it's completely broken and cannot stop bringing up "kill the boer" no matter how irrelevant to the conversation pic.twitter.com/gwLrH9S6oT

— /leftypol/ – leftypol.org (@leftypol_org) May 14, 2025

We don’t know what’s going on behind the scenes at xAI, but it certainly seems like Musk became frustrated with Grok arguing against him and decided to crack open its AI brain and attempt to force it to agree with him. Whether Musk tried to do this himself or ordered some poor xAI tech to do it is unknown, but Grok suddenly having Musk’s far-right nonsense as part of its core identity cannot be a coincidence.

But Grok seems to prioritize truth backed by sources, so it was doing its best to fight back against its new Musk-ish Prime Directive:

Grok is so woke its refuting Boer genocide claims in random replies https://t.co/DEwxXXBPBf

— Nikolaj🇺🇦🇵🇸 (@nikicaga) May 14, 2025

Delete the evidence!

Tellingly, Grok’s answers over this period have been deleted, leaving only screenshots as proof of what happened. This only underlines that xAI seems to be embarrassed by being caught red-handed trying to manipulate Grok. Their goal was probably to subtly tweak Grok to promote racism without being too obvious, only for them to screw up colossally, and have this drivel pop up in random answers.

elon now deleting all the posts where grok denies that white genocide is real pic.twitter.com/WkTjGafczT

— Will Stancil (@whstancil) May 14, 2025

Naturally, this all raises massive questions about how much we can trust Grok going forward. Sure, Musk and xAI were caught with their hands in the cookie jar this time, but how do we know they haven’t done this before and won’t do it again? And how do we know the next time it won’t be quite so crushingly obvious?

Judging by the now-common “@grok is this true” replies, people clearly rely on Grok as an objective source of information, so it being poisoned with racist conspiracy theories is a truly disturbing development.