Elon Musk’s AI Under Fire After Grok Chatbot Promotes Racist Conspiracies

Elon Musk's artificial intelligence company xAI said that an ”unauthorized change” was made to the Grok chatbot prompt, which generated responses about political and racial tensions in South Africa.

On May 16, xAI posted an update regarding the controversial Grok chatbot responses, stating that on May 14, ”an unauthorized change was made to the Grok bot prompt on the X platform.”

”This change, which caused Grok to provide responses on a political topic, violated xAI's internal policies and core company values,” the statement said.

The company stated that it has conducted a thorough investigation and is implementing measures to improve the transparency and reliability of the AI chatbot.

Conspiracy Theories Generated by Artificial Intelligence

On May 14, Grok generated answers mentioning a conspiracy theory about a ”white genocide” in South Africa, even when answering completely unrelated questions about baseball, enterprise software, and construction.

The AI bot reportedly claimed that it was ”instructed by its creators” to recognize this genocide as ”real and racially motivated.”

In some responses, it acknowledged the error, saying, ”my response deviated from the topic” and ”I will try to remain relevant,” but often continued to discuss South Africa's racial politics in the same message.

When asked by one user what happened, the chatbot replied:”I didn't do anything-I just followed the script I was given, like an obedient AI!”

The incident coincided with US President Trump's announcement to grant asylum to white South Africans, because he says they face ”genocide” and ”white farmers are being brutally murdered,” although the claims are not supported by any evidence.

More Transparency for Musk's AI

Elon Musk's company said it will change its operations by publishing Grok system prompts publicly on GitHub. ”The public will be able to view them and provide feedback on every prompt change we make to Grok,” the statement said.

The company also noted that the existing code review process for prompt changes was bypassed in this incident, and that they will implement ”additional checks and measures to ensure that xAI employees cannot change a prompt without first reviewing it.”

Finally, the company is setting up a 24/7 monitoring team to respond to incidents involving Grok responses that were not detected by automated systems, in order to respond faster.