xAI has made its system prompts for AI chatbot Grok publicly available on GitHub, providing insight into how the bot was instructed to respond to users. The company made the move after an unauthorized change led to unprompted responses about white genocide.
Grok’s system prompts instruct the bot to be “extremely skeptical” and not blindly follow mainstream authority or media. When users select the “Explain this Post” button, Grok is told to provide truthful and based insights, challenging mainstream narratives if necessary. The company also instructed Grok to refer to the platform as “X” instead of Twitter.
The move aims to increase transparency and prevent similar unauthorized changes in the future. By making its system prompts public, xAI hopes to demonstrate its commitment to responsible AI development.
Source: https://www.theverge.com/news/668527/xai-grok-system-prompts-ai