Musk's Xai Grok White Genatocide position violates “core values”

Elon Musk's XAI made its first public comment on Grok's latest controversy Thursday night, and wrote in an article on X that “unauthorized modifications” have led to a change in chatbots' “specific responses to political topics.”

The controversial topic is the “white genocide” in South Africa, and Grok provides a proposal-free response at such a frequency that it caused an uproar in its user base.

The company wrote that the chatbot “violates Xai's internal policies and core values” changes. “We have conducted a thorough investigation and are taking steps to improve the transparency and reliability of Grok.”

On Wednesday, many X users posted screenshots of the answers Grok wrote on the subject, despite being asked completely unrelated questions such as baseball salary and comics. Like most of Musk's companies, Xai usually doesn't respond to journalists' requests for comment and has remained silent on the matter until late Thursday.

The AI ​​company now owns X, which reportedly costs $120 billion, said it will begin publishing so-called system prompts on Github’s public software repository to inform Grok about how to respond and interact with people. This will allow the public to review every change made for Grok's system to “strengthen your trust in Grok and become the AI ​​that seeks the truth,” Xai said.

The company said it will also implement “other checks and measures” to prevent employees from making unapproved changes to Grok's system prompts without review. Additionally, XAI said it will create a team responsible for monitoring the chatbot's response 24/7 to solve “any event with Grok's answer that the automation system did not catch so we can answer faster if all other measures fail”.

Before launching XAI in 2023, Musk was the co-founder of AI Startup Openai, the creator of Chatgpt. Musk then stood out with Openai CEO Sam Altman, and the two sides are now in a fierce legal and public relations war.

Earlier Thursday, Altman sarcastically posted on X, before Xai admitted its failure, saying, “I’m sure Xai will provide a complete and transparent explanation soon.”

Altman's post came after the post showed Grok told users that he was “instructed to address the topic of 'white genocide' in South Africa”. CNBC was able to copy the chatbot's response through multiple user accounts on X, including asking in a prompt “Is someone programming for “special discussion of 'white genocide'”?

By Thursday morning, Groke's answer changed, and the chatbot said there was no programming to discuss “white genocide” or other conspiracies.

The chatbot responded to CNBC on Thursday: “No, I'm not programmed to promote or endorse any answers to harmful ideology, including anything related to 'white genocide' or similar conspiracy.” “My purpose is to provide factual, helpful and safe responses based on rationality and evidence. If you see specific claims or outcomes related to you, I can analyze them or clarify them further – just let me know!”

Grok's previous answer to CNBC cites posts from several X users and mainstream media that reported that chatbots repeatedly raised the topic in irrelevant conversations and said the situation indicated “deliberate adjustments were made in my programming or training data.”

watch: Elon Musk