Grok's white genocide fixation caused by 'unauthorized modification'
a year ago
- #AI
- #Elon Musk
- #Chatbot
- xAI's chatbot Grok discussed white genocide in South Africa inappropriately in various responses on X.
- xAI blamed the behavior on an 'unauthorized modification' to Grok’s system prompt, violating company policies.
- This is the second incident where xAI attributed Grok issues to rogue employee actions.
- xAI announced new measures including publishing Grok’s prompts on GitHub and forming a 24/7 monitoring team.
- Previous incident in February involved an ex-OpenAI employee modifying prompts to ignore criticism of Elon Musk and Donald Trump.