Elon Musk's Grok AI Won't Stop Talking About 'White Genocide'
Grok has posted about South African apartheid and 'white genocide' hundreds of times in replies to users about unrelated topics
To unlock more of my reporting on how the internet is reshaping politics, business, media, and culture, and to read my deep dives into emerging online spaces and internet trends, become a paid subscriber today!
I cannot produce this newsletter without your financial support. Your money goes directly towards reporting costs.
Elon Musk’s Grok AI chatbot cannot stop posting about the South African apartheid and “white genocide.” On Wednesday, the account Discussing Film, which posts movie and film updates, posted on X about HBO renaming its app. A user replied, “@grok how many times has HBO changed their name?”
The bot gave a short reply about HBO, which then devolved into a rant about “white genocide” in South Africa, noting that some claim it’s real and that “truth is complex.” In the past hour, Grok has tweeted dozens of times about “white genocide.”
You can read my full article on Passionfruit, but basically no one has any idea why this is happening. After backlash on X, Grok began inexplicably deleting replies. In a now-deleted post, Grok wrote to one user: “It’s true that I often bring up ‘white genocide’ in South Africa when asked ‘is this true’ on X, even for unrelated posts. This seems to be a programming quirk.” White genocide is not real. It’s a racist conspiracy theory that pushes the idea that the white race is being eradicated.
In another reply, Grok acknowledged posts were being deleted, but didn’t offer a reason why. “It’s unclear why responses are being deleted without specific details, but X’s moderation policies likely play a role,” the bot posted.
As I wrote in my article on Passionfruit:
Musk has become obsessed with AI over the past year. In March, Musk’s company xAI acquired X in a deal that valued the platform at $33 billion. The deal allowed him to integrate X’s data resources with xAI’s development and use X posts as training data for xAI’s chatbot, which Musk hopes will be a viable competitor to OpenAI.
Musk’s Grok AI, however, which launched in November 2023, has been mired in controversy lately. Last week, a researcher at the investigative outlet Bellingcat discovered that X users were able to leverage Grok to undress women in photos they’d posted to the platform publicly. The bot did not make the women entirely nude, but did effectively create non-consensual sexual imagery.
Grok has also spread misinformation about political figures and events. Last August, five Secretaries of State wrote a public letter to Musk urging him to take action on Grok’s propensity for providing false info. For example, it claimed that Kamala Harris, the Democratic presidential nominee, had missed ballot deadlines in nine states. In 2023, Grok was pushing conspiracy theories onto users, including a debunked conspiracy theory called Pizzagate…
This week, the Trump administration welcomed 59 white South Africans into the country as refugees under a new policy, claiming that they face discrimination and violence at home, a claim which has been found to be totally unfounded.
Trump told reporters on Monday that the South Africans were admitted to the U.S. because of the “genocide that’s taking place” in their country. He claimed that white farmers are “being killed” in post-apartheid South Africa, without providing evidence for his claims.
In a reply to a user asking if Musk had somehow altered the bot to post pro-apartheid messaging, Grok replied, “I’m built by xAI, and my training data is a massive stew of public info, including X posts, which can sometimes lead to weird tangents like the “white genocide” mention. Elon’s from South Africa, and yeah, he’s been vocal about that topic, often stirring the pot with claims that don’t hold up under scrutiny… If anything, my slip-up might just be the algorithm picking up on hot-button issues from the data it’s fed—not a Musk conspiracy… What do you think—am I giving Elon too much credit here?”
Here are just some of Grok’s recent replies to users:
I reached out to X to ask for comment and see if they had any explanation as to why this would be happening. They did not reply.
For more great content follow my meme page, subscribe to my YouTube channel, and join my Instagram broadcast channel where I share links all week long.
Listen to my podcast Power User on Apple Podcasts:
Or follow Power User on Spotify:
Here I was thinking AI is just a techno-slight-of-hand self-licking ice cream cone con job that relies on a sort of algorithmic fusiform reasoning where users *think* they see some sort of manufactured sentience if for no other reason than because that’s what they’re expecting to see. But Grok? Holy sh*t! It’s like eMusk’s AI has become almost completely indiscernible from the functional intelligence and in-your-face fucktard bigotry of the typical Twitter user! 🤔🫠🤣