xAI says Grok’s posts on South Africa racial politics were due to “unauthorised modification”

xAI has commented on the controversy around their AI chatbot Grok repeatedly addressing the claims about “white genocide” in South Africa unprompted in its responses. The Elon Musk-led AI firm said on X that an “unauthorised modification” had been made to Grok’s system prompt on May 14 (Wednesday) which “directed it to provide a specific response on a political topic, violating xAI’s internal policies and core values.” 

The statement said that they had conducted an investigation and would implement measures to make the chatbot more transparent and reliable. 

Moving forward, xAI promised to publish Grok system prompts openly on GitHub so the public can review them easily and give feedback. There will also be “additional checks and measures to ensure xAI employees can’t modify the prompt without review,” xAI said. 

Additionally, a 24/7 monitoring team will be available to respond to similar incidents. 

For the past couple of days, several Grok users shared screenshots on X of responses that spoke about the subject despite being asked queries that were completely unrelated. The chatbot then told users that it had been “instructed to address the theory of ‘white genocide’ in South Africa.”

OpenAI CEO Sam Altman also remarked sarcastically on the incident before xAI’s statement saying, “I’m sure xAI will provide a full and transparent explanation soon.”