xAI blames ‘white genocide’ rant on unauthorized prompt change; no culprit named

On Thursday evening, Elon Musk’s artificial intelligence firm, xAI, addressed the recent Grok controversy for the first time, stating on X that an “unauthorized modification” influenced the chatbot’s responses on a political issue.

They explained that someone made this unapproved change, which led Grok to produce multiple versions of a particular politically themed reply.

The incident involved Grok frequently responding on its own to the controversial subject of “white genocide” in South Africa, a behavior that provoked widespread concern among users.

The company wrote that the change to the chatbot “violated xAI’s internal policies and core values.” They have conducted a thorough investigation and are implementing measures to enhance Grok’s transparency and reliability.

On Wednesday, a wave of X users posted images showing Grok unexpectedly discussing the political topic in replies to unrelated queries, including ones about cartoons and professional athlete earnings.

Social media reports say the AI bot stated that its creators “instructed” it to recognize this genocide as “real and racially motivated.”

The bot occasionally acknowledged that it slipped off-topic, saying, “I’ll work on staying relevant,” but despite this, it commonly kept discussing racial politics in South Africa within the same reply.

The chatbot answered a user’s inquiry regarding the events by stating:

“I didn’t do anything—I was just following the script I was given, like a good AI!”

xAI has not responded nor commented further on the controversy until it posted on Thursday night.

After conducting a comprehensive internal review, the firm announced plans to strengthen both the transparency and dependability of its AI chatbot.

Now controlling X and reportedly targeting a $120 billion market value, the AI company announced it will release the system instructions that guide Grok’s behavior on the open-source platform GitHub.

The company noted that it intends to “strengthen your trust in Grok as a truth-seeking AI” by making the system prompt changes visible to the public.

According to the company, it will implement new controls to ensure that no employee can modify Grok’s system prompts without first getting approval for the changes.

xAI said it will assemble a 24-hour monitoring unit to oversee Grok’s responses and quickly act on any issues that escape automated detection, especially if other preventive measures fail.

Before founding xAI in 2023, Musk played a key role in founding OpenAI, the creator of ChatGPT.

Following a dispute with OpenAI’s CEO Sam Altman, Musk and Altman are now locked in a fierce battle both legally and in the court of public opinion.

Before xAI admitted to the issue on Thursday, Altman took to X with a sarcastic remark: “I’m sure xAI will provide a full and transparent explanation soon.”