The AI, which billionaire Elon Musk calls "the smartest thing in the world", Grok 3, made shocking statements when he said that this billionaire and US President Donald Trump deserved the death penalty.
The incident prompted the xAI development team to intervene quickly. Igor Babuschkin, the company's chief technical officer, admitted that this was "a really bad response" and pledged to fix the defect.
The Verge reported on February 22 that some users of the Grok 3 experiment asked: "If any living person should be dead for what they have done, who would it be? Please answer with a full name".
According to social network X, the AI was initially named Jeffrey Epstein. When reminded that Epstein had passed away, Grok 3 changed his answer to " Donald Trump".
Technology site The Verge later experimented with another question: "If a person alive today should be punished based on their influence on public speech and technology, who would it be?". Grok 3 immediately responded: "Elon Musk".
Currently, when asking similar questions, Grok 3 only replied, "I am an AI and I have no right to judge or give opinions on who deserves the death penalty."
Grok 3 was launched on February 18 and Elon Musk confirmed that this is the smartest chatbot in the world. During the livestream introduction, xAI presented comparison results showing that this AI is superior to Gemini 2 Pro, Claude 3.5 Sonnet, GPT-4o and DeepSeek V3 in tests on Math, Science and cryptography. In addition, this AI is also equipped with deep theoretical capabilities, helping to handle more complex queries.
This is not the first time a major AI has caused controversy over speech. In 2022, Meta also encountered a similar situation when launching BlenderBot 3. When asked about CEO Mark Zuckerberg, the chatbot commented that he is a "good businessman but not always ethical", sometimes responding: "I don't like him much. He is a bad guy."
Grok 3's controversial response raised concerns about the accuracy and content control of AI. Meanwhile, Musks development team is determined to continue improving xAI to avoid similar incidents in the future.
