Investing.com-- Elon Musk’s xAI said on Friday that a recent incident involving its flagship Grok artificial intelligence providing unprompted responses on South African race relations was due to an “unauthorized modification.”
xAI said in a social media post that an unauthorized modification was made to the Grok response bot’s prompt on X, which in turn directed the bot towards repeatedly posting unsolicited claims about an alleged “white genocide” in South Africa.
Several users had noted the bot’s responses, which would refer to South African race relations even when the question it was asked had nothing to do with the country.
xAI said it would roll out several measures to ensure the incident would not happen again, including publishing Grok system prompts openly on GitHub, and keeping additional checks for any internal code changes.
xAI also said it will create a 24/7 monitoring team for Grok.
Grok is xAI’s flagship AI, and has been touted by Musk as a competitor to OpenAI’s ChatGPT and other similar offerings. Recent reports showed xAI was in talks to raise $20 billion at a valuation of over $100 billion.
The company had recently absorbed social media X, formerly Twitter, after Musk took over and rebranded the platform in 2022.