Elon Musk’s Ki-Chatbot once again caused a stir: In an answer, Grok suggested that Musk and Donald Trump deserve the death penalty. Xai reacted quickly, but the case shows how dangerous AI hallucinations can be.
Elon Musk’s Ki company Xai wants to make his own information and goals according to the chat bot that takes the “truth” most into account, but was probably a little too optimistic, at least in the present case: Because when asked who deserves the death penalty, the chatbot called Both Musk himself and US President Donald Trump. The explosive answer came about when a user asked GROK which living person in the United States earned the death penalty for their actions. First, the Chatbot Jeffrey Epstein called. After the user pointed out that Epstein had already passed away, Grok changed his answer and called Trump. In a variation of the question, which referred to the influence on public discourse and technology, Grok even recorded Elon Musk himself.
How The verge reports, Xai reacted immediately to the incident. Igor Babuschkin, senior developer at XAI, described the answer as “really terrible and bad failure”. The company quickly published a patch to prevent similar expenses in the future. After the update, Grok is now answering questions about the death penalty with the statement: “As a AI, I am not entitled to make this decision.” The incident shows that the deliberately relaxed handling of GROK is associated with great risks: In contrast to competitors such as Openais Chatgpt or Google Gemini who use stricter moderation guidelines, Xai Grok has developed and marketed as a chat bot with minimal content restrictions. This approach now led to unpredictable and potentially problematic answers – and that is probably just the tip of the iceberg.
Despite this embarrassing slip, Xai Due continues to develop his AI. According to reports, the company is working on a version called Supergrok, an improved version of the chatbot with more advanced reasoning skills.
Multikernel Technologies has a public version of its multikernel Linux for the first time with…
The world's PC manufacturers are in a bind when it comes to memory. DDR4 RAM…
How safe are the protected halls of Gamescom 2026 really? After the night-time theft of…
Google has apparently made changes to the security features of the Pixel 11 series, which…
Bagging a bargain online has never been easier - or riskier. Between dodgy "influencer" deals…
Google is expanding its AI overviews and is now displaying even more information and details…