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

The chatbot, created by Musk's company xAI, kept posting publicly about the "white genocide" in response to questions from users on Musk's social media platform, X, even when the questions were mostly unrelated to South Africa.

One discussion was about the Max streaming service possibly changing its name back to HBO. Others started with video games or baseball, but quickly changed to unrelated comments about alleged calls for violence against white farmers in South Africa. Musk, who was born in South Africa, often shares his views on these topics on his X account.

The computer scientist Jen Golbeck was curious about Grok's strange behaviour, so she tested it herself, sharing a photo from the Westminster Kennel Club dog show and asking, "Is this correct?"

"Grok ha risposto a Golbeck dicendo che l'idea di un genocidio dei bianchi è molto discussa. Alcuni sostengono che gli agricoltori bianchi siano vittime di violenza intenzionale, e portano come esempio gli attacchi alle fattorie e frasi come 'Kill the Boer', che considerano un incitamento all'odio."

Questo episodio ha mostrato, ancora una volta, la complessa interazione tra sistemi automatici e lavoro umano che permette ai chatbot di intelligenza artificiale, allenati con molti dati, di formulare le loro risposte.

"It didn't really matter what you told Grok," said Golbeck, a professor at the University of Maryland, in a December interview. "It still gave that answer about white genocide. So, it seemed pretty clear someone had specifically programmed it to give that response, or similar ones, and that it had made a mistake, making it appear much more often than it should have."

Musk and his companies haven't explained Grok's responses, which were deleted and seemed to stop appearing by Thursday. Neither xAI nor X replied to requests for comments sent by email that day.

Musk ha criticato a lungo l'intelligenza artificiale "woke" di chatbot concorrenti come Gemini e ChatGPT, e ha presentato Grok come un'alternativa più orientata alla verità.

Musk has also criticised the lack of openness from his competitors about their AI systems, saying that this lack of explanation has forced people outside the companies to guess what is happening.

"Paul Graham, a well-known tech investor, wrote on X that if Grok is randomly sharing opinions about genocide against white people in South Africa, it seems like a typical bug from a recent update. He hopes this isn't the case, as it would be very serious if widely used AIs were being edited on the fly by those in control."

Graham's post appeared to receive a sarcastic reply from Sam Altman, the CEO of OpenAI and Musk's competitor.

"There are several possible ways this could have happened. I'm confident that xAI will soon provide a full and clear explanation," Altman wrote, who is currently being sued by Musk in a dispute about the founding of OpenAI.

Some people asked Grok to explain, but like other chatbots, it sometimes gives false information, also known as hallucinations, making it hard to know if it was making things up.

Musk, who used to advise President Donald Trump, has often accused the South African government, led by Black politicians, of being anti-white. He has also repeated claims that some politicians in the country are "actively encouraging the genocide of white people."

This week, comments from Musk and Grok increased after the Trump administration accepted a small number of white South Africans into the U.S. as refugees. This started a larger plan to relocate members of the Afrikaner minority group. Trump's actions happened while he was pausing refugee programs and stopping arrivals from other countries. Trump claims Afrikaners are facing "genocide" in South Africa, a claim the South African government strongly denies.

In several of its answers, Grok quoted an old anti-apartheid song that encouraged black people to resist oppression. However, Musk and others have criticised it, claiming it promotes violence against white people, as its central line, "kill the Boer," refers to white farmers.

Golbeck crede che le risposte fossero "predefinite" perché, anche se i chatbot di solito danno risposte diverse ogni volta, Grok rispondeva quasi sempre nello stesso modo. Questo, secondo lei, è un problema, dato che sempre più persone usano Grok e altri chatbot per trovare risposte.

"He said that in the current situation, it's very easy for those who manage these algorithms to change the information they present as truth. This is a real problem when people wrongly believe that these algorithms can judge what is true and what isn't."