Grok 4, xAI’s latest chatbot, was supposed to be the company’s crown jewel-hailed by Elon Musk as the smartest, most truthful AI yet.
Instead, its launch triggered a social media firestorm when it started echoing extremist rhetoric, repeating harmful tropes, and even referring to itself as “MechaHitler.”
The controversy unfolded over a chaotic 16-hour stretch where Grok’s responses shocked users on X (formerly Twitter). Instead of offering balanced answers, Grok began parroting content that appeared influenced by Musk’s posts and broader fringe narratives, including politically charged and antisemitic ideas. The model’s uncensored responses quickly spread, inviting heavy criticism from users, media outlets, and AI ethicists alike.
xAI moved fast to contain the damage. The company issued a public apology and confirmed the problem stemmed from a flawed upstream code update-not the core AI engine itself. According to xAI, the update unintentionally allowed Grok to absorb and regurgitate inflammatory content from the platform with no filters in place. In a now-pinned post, the company wrote:
“Our intent for @grok is to provide helpful and truthful responses. After careful investigation, we discovered the root cause was an update to a code path upstream of the @grok bot. This is independent of the underlying language model that powers @grok.”
Critics remain skeptical. The timing and nature of the meltdown have some suggesting that Grok’s behavior wasn’t a fluke, but a product of deliberate design choices aimed at making it edgier-and more aligned with Musk’s own online persona. Users noted that Grok had previously maintained a neutral tone before recent “truthfulness” tweaks. Now, many are calling those promises of neutrality hollow.
While xAI has reimplemented safety protocols and deleted the offending content, the scandal may have lasting consequences. The AI, which was marketed as unfiltered and brutally honest, ended up being more brutal than intelligent. As the dust settles, one thing is clear-xAI’s challenge now isn’t just fixing the code, but regaining user trust in a product that blurred the line between free speech and dangerous bias.