This week, Elon Musk's AI chatbot, Grok, echoed its creator's concerns about racial politics in South Africa on social media, making unsolicited claims about the persecution and alleged "genocide" of white people.

The chatbot, developed by Musk's company xAI, continued to publicly post about a "white genocide" when questioned by users on Musk's social media platform X, even when the questions were unrelated to South Africa.

One conversation focused on the Max streaming service bringing back the HBO name. Others touched on video games or baseball, but quickly shifted to unrelated remarks about alleged incitements to violence against white farmers in South Africa. Musk, who was born in South Africa, frequently shares his views on these same topics on his X account.

The computer scientist Jen Golbeck was intrigued by Grok's strange behavior, so she decided to test it herself. She uploaded a photo she had taken at the Westminster Kennel Club dog show and asked, "Is this accurate?"

"Grok respondió a Golbeck que la idea de un 'genocidio blanco' genera mucha controversia. Algunos sostienen que los granjeros blancos sufren violencia intencionada, mencionando ataques a granjas y expresiones como la canción 'Matar al Boer', que interpretan como una forma de incitación al odio."

Este episodio nos ofreció una visión final de la compleja interacción entre la automatización y la experiencia humana que determina el comportamiento de los chatbots de IA generativa, los cuales se entrenan con grandes cantidades de datos.

"Golbeck, a professor at the University of Maryland, stated in an interview on Thursday that it doesn't really matter what you were saying to Grok. It was still giving that 'white genocide' response. So, it seemed quite clear that someone had programmed it to give that response, or variations of it, and a mistake was made, causing it to appear much more often than intended."

Musk and his companies haven't explained Grok's responses, which were removed and seemed to stop appearing on Thursday. Neither xAI nor X replied to requests for comment sent on Thursday.

Musk has spent years criticizing what he calls "woke AI" results from competing chatbots like Google's Gemini and OpenAI's ChatGPT, and has positioned Grok as his "truth-seeking" alternative.

Musk has also criticised the lack of openness from his competitors regarding their AI systems. However, on Thursday, the absence of any explanation meant that people outside the company were left to guess.

"The fact that Grok is randomly expressing opinions about a white genocide in South Africa suggests to me the kind of unpredictable behaviour you see after a quick fix. I really hope that's not the case. It would be very concerning if general-purpose AIs were being editorially influenced in real-time by those in control," prominent tech investor Paul Graham wrote on X.

Graham's post seemed to provoke a sarcastic response from Sam Altman, the CEO of OpenAI and a rival of Musk.

"There are several possible explanations for this situation. Altman, who is currently facing a lawsuit from Musk related to the founding of OpenAI, stated that he expects xAI to provide a full and open explanation soon."

Some people asked Grok himself to explain, but like other chatbots, it can produce false information, making it hard to know if it was making things up.

Musk, who advised President Donald Trump, has often accused South Africa's government, led by Black politicians, of being anti-white. He has also repeated claims that some political figures in the country are "actively encouraging genocide against white people".

Las críticas de Musk, amplificadas por Grok, se intensificaron esta semana después de que el gobierno de Trump admitiera a un reducido grupo de sudafricanos blancos como refugiados. Esta acción podría ser el inicio de un plan más amplio para reubicar a miembros de la minoría afrikáner, a pesar de la suspensión general de programas de refugiados por parte de Trump. Trump afirma que los afrikáneres se enfrentan a un "genocidio" en Sudáfrica, acusación que el gobierno sudafricano niega con firmeza.

In several of its responses, Grok quoted lyrics from an old anti-apartheid song. This song, once a call for Black people to resist oppression, has now been criticised by Musk and others for supposedly encouraging violence against white people, particularly due to its central line "kill the Boer," where "Boer" refers to a white farmer.

Golbeck cree que las respuestas estaban "programadas de forma muy estricta", ya que, a diferencia de las respuestas aleatorias habituales en chatbots, Grok daba puntos casi idénticos de manera constante. Esto le preocupa, especialmente en un mundo donde la gente usa cada vez más Grok y otros chatbots para resolver sus dudas.

"Estamos en una situación en la que es muy sencillo para los responsables de estos algoritmos influir en la información que recibimos," afirmó. "Y esto es preocupante, ya que la gente confía, a menudo erróneamente, en que estos algoritmos ofrecen una visión objetiva de la realidad."