Musk's xAI updates Grok chatbot after 'white genocide' comments

(Reuters) -Elon Musk's xAI responded to wide reports that its Grok chatbot made claims astir a genocide against achromatic citizens successful South Africa, saying determination had been an unauthorized alteration to the artificial quality bot. In a station connected X connected Thursday, xAI said it would update the strategy to code the problem.

Issues of governmental biases, hatred code and accuracy of AI chatbots person been a interest since astatine slightest the motorboat of OpenAI's ChatGPT successful 2022.

XAI said that aboriginal connected Wednesday the unauthorized alteration was made to Grok's effect software, circumventing the mean reappraisal process.

"This change, which directed Grok to supply a circumstantial effect connected a governmental topic, violated xAI's interior policies and halfway values," xAI said.

Some X users connected Wednesday noticed that Grok brought up the taxable of "white genocide" successful South Africa successful unrelated discussions astir different matters, sharing screenshots of their exchanges.

Critics of a onshore expropriation argumentation successful South Africa, including achromatic South African-born Musk, person called it racist against whites. The South Africa authorities says that determination is nary grounds of persecution and that claims by U.S. President Donald Trump and others of "genocide" are unfounded.

In effect to the incidental this week, xAI said it would openly people Grok's strategy prompts connected GitHub, wherever the nationalist tin presumption and supply feedback connected each punctual alteration made to the chatbot.

It would enactment successful spot a round-the-clock monitoring squad to respond to incidents with Grok's answers that are not caught by automated systems, it added.

(Reporting by Zaheer Kachwala successful Bengaluru; Editing by Peter Henderson and Rosalba O'Brien)