This week, Elon Musk's AI chatbot "Grok," much like its creator, became obsessed with South African racial politics on social media, posting unnecessary claims about the persecution and "genocide" of white people.

The chatbot created by Musk's company, xAI, responded to user questions on his social media platform X by repeatedly posting about "white genocide," a topic largely unrelated to South Africa.

In one exchange, they talked about the possibility of the streaming service Max bringing back the HBO name. Another discussion, about video games and baseball, quickly changed to irrelevant comments that seemed to encourage violence against white farmers in South Africa. Musk, who was born in South Africa, often shares his opinions on similar topics on his X account.

Intrigued by Grok's unusual behavior, computer scientist Jen Golbeck decided to investigate it herself. She shared a photo from the Westminster Kennel Club Dog Show, questioning, "Is this even real?"

Grok began his reply to Golbeck by stating that the idea of "white genocide" is highly controversial. He added that some people claim white farmers are facing targeted violence, and they see rhetoric, such as the song "Kill the Boer," as incitement.

This episode is a recent example of how generative AI chatbots, trained on huge amounts of data, shape their responses through a complex mix of automation and human design.

"Professor Golbeck from the University of Maryland explained in an interview on Thursday that regardless of the question asked to Grok, the response was always "white genocide." She believes that someone deliberately programmed this answer, or a similar version, into the system, and a malfunction is causing it to appear more often than it should."

Neither Musk nor his companies have explained Grok's responses, which were removed and stopped spreading by Thursday. As of Thursday, neither xAI nor X responded to requests for comment via email.

Musk has often criticised what he calls "woke AI," referring to the outputs from rival chatbots like Google's Gemini and OpenAI's ChatGPT, and presents Grok as an alternative that aims to be "maximally truth-seeking."

Musk has criticised other AI companies for not being open enough about their systems. However, he didn't explain anything on Thursday, leaving outsiders to guess what was happening.

Paul Graham, a well-known tech investor, posted on X that Grok's random comments about white genocide in South Africa seem like a bug caused by a recent update. He added that if this is the case, it's a serious issue, as it's very problematic for those in charge to be able to edit a widely used AI on the spot.

Sam Altman, the CEO of OpenAI and a rival of Musk, responded to Graham's post with some irony, saying that there could be many explanations for what happened, and he hoped xAI would provide a full and clear account soon.

Some people asked Grok for an explanation, but like other chatbots, it can create false information, making it difficult to know what is real and what is made up.

Musk, who advises President Donald Trump, often accuses South Africa's Black government of being anti-white. He also claims that some of the country's politicians are "actively pushing for white genocide."

Comments from Musk, and Grok, have made the situation worse this week after the Trump administration brought a small number of white South Africans to the US as refugees on Monday. This happened while Trump has stopped the refugee program and blocked arrivals from other areas, raising questions about a bigger plan to move Afrikaner people. Trump says Afrikaners are facing "genocide" in their country, but the South African government strongly disagrees.

In many responses, Grok quoted lyrics from an old anti-apartheid song. The song encourages Black people to resist oppression, but it's now accused by figures like Musk of promoting violence against white people. The central lyric, "Kill the Boer," refers to white farmers, with "Boer" being the term used for them.

Golbeck believes the responses are "hard-coded" because, unlike the usually random output of chatbots, Grok consistently focused on the same arguments. This is very worrying in a world that relies on Grok and other AI chatbots for answers.

She argued that the administrators of these algorithms could easily control what is presented as the truth, and it's deeply concerning that people trust these algorithms to decide what is true and false.