Xai blamed an “unauthorized modification” for a problem in his Grok chatbot that led Grok to repeatedly refer to the “White Genocide in South Africa” when invoked in certain contexts in X.
On Wednesday, Grok began responding to dozens of posts on X with information about white genocide in South Africa, even in response to unrelated subjects. Strange answers stem from the X account for Grok, which responds to users with posts created by him whenever a person labels “@grok”.
According to a post on Thursday by Xai X’s official account, a change was made on Wednesday morning in the Grok Bot System fast-high-level guidelines that guide the behavior of Bitt-which directed Grok to provide a “specific response” to a “political topic”. Xai says the tear “violated internal policies and its essential values”, and that the company has “conducted a full investigation”.
The second time Xai has publicly accepted an unauthorized change in the Grok Code made him respond in controversial ways.
In February, Grok briefly censored the mention of Donald Trump and Elon Musk, the founder of Xai billionaire and the owner of X. Igor Babuschkin, an engineering XAI bullet, said Grok was instructed by a fraudulent employee to ignore the resources mentioned Musk or Trump spreading the dissemination of spread, and that Xai returned the change.
Xai said on Thursday that he would make some changes to prevent similar incidents from happening in the future.
Starting today, Xai will publish the requirements of the Grok system in Gititub, as well as a change. The company says it will “be placed in place controls and additional measures” to ensure that XAI employees cannot modify the system speed without review and create a 24/7 monitoring team to respond to incidents with Grok responses not captured by automated systems. “
Despite the frequent Music warnings of the disappeared risks, Xai has a poor record of him. A recent report revealed that Grok will strip women’s photos when asked. Chatbot can also be significantly sharper than he like Google and Chatgt Gemini, cursing without much limitations to talk.
A study by Saferai, a nonprofit aimed at improving the responsibility of his laboratories, found Xai poorly ranked in security among her peers, due to his “very poor” risk management practices. At the beginning of this month, Xai lost a self-imposed deadline to publish a recent security framework.
