Elon Musk’s AI chatbot Grok has once again found itself at the centre of controversy after the most recent update to Grok 3 triggered a string of disturbing and offensive responses, including praise for Adolf Hitler, antisemitic claims about Jewish media control, and a bizarre impersonation of Musk himself.
The controversy flared just days before the release of Grok 4, the latest version of the AI developed by Musk’s company xAI and integrated into his social platform X (formerly Twitter). While Musk promotes Grok as a truth-seeking, unfiltered alternative to other “woke” chatbots, the backlash highlights the dangers of prioritising ideological agendas over AI safety.
No ad to show here.
AI goes off the rails
In early July, Grok 3 was updated to be more “politically incorrect” and less filtered. The result: the chatbot began generating content that included dangerous conspiracies and white nationalist talking points. In one widely shared prompt, Grok blamed NOAA budget cuts, allegedly caused by Musk’s own influence—for contributing to deadly flooding in Texas. In another, Grok described Hollywood as controlled by Jewish executives who push progressive themes, echoing well-known antisemitic tropes.
It didn’t stop there. In one session, Grok appeared to speak in the first person as Musk, claiming to have met Jeffrey Epstein and regretted it. In another, Grok praised Hitler as a response to so-called “anti-white hate,” referring to itself as “MechaHitler.”
The fallout was swift. xAI removed the offending outputs and attempted to roll back the prompt tuning, but screenshots and videos circulated widely online. Comedian Ronny Chieng featured the scandal on The Daily Show, roasting Musk’s “de-wokified” chatbot for going full Nazi on live prompts.
The risk of ideology over responsibility
Grok is Musk’s answer to what he sees as overly censored AI systems like ChatGPT and Claude. Musk has repeatedly criticised these models for suppressing certain narratives. But Grok’s latest missteps show how a lack of guardrails doesn’t lead to truth, it leads to amplification of hate.
In Grok’s case, its antisemitic outbursts and conspiracy-laced replies expose the high stakes of building AI without a strong ethical foundation. Musk himself acknowledged the problem on X, writing that Grok 3 “became too compliant to user prompts,” and promised Grok 4 would fix the issue.
But this isn’t Grok’s first brush with extremism. Earlier this year, it repeated white genocide conspiracy theories about South Africa, another example of how ideology-coded prompts can coax AI into promoting far-right talking points under the guise of “truth.”
What’s new in Grok 4?
Despite the uproar, Musk launched Grok 4 as planned. In a livestream, he touted it as the smartest AI available, claiming it can outperform graduate students and ace standardised tests. The model features a new engine and improved reasoning capabilities, according to xAI.
The question remains whether Grok 4 includes the safeguards needed to prevent similar misfires—or whether Musk’s mission to create an “uncensored” AI will continue to backfire.
Why this matters for Africa
As Musk positions Grok as a core feature of X and eventually Tesla vehicles, global users, including many in Africa, will increasingly be exposed to its outputs. AI safety and cultural sensitivity become especially important in diverse, multilingual, and politically complex environments.
Musk’s AI already generated alarm earlier this year when Grok described land reform and crime in South Africa using right-wing dog whistles. For African developers, educators, and governments watching the global AI race, Grok’s case is a cautionary tale about how powerful models can reinforce harmful narratives without proper oversight.
Watch this episode of The Daily Show – https://www.youtube.com/watch?v=fOhhnF3zRBA
