Grok, the AI chatbot developed by Elon Musk’s xAI, recently altered its response protocols to block results that mention Elon Musk or Donald Trump as spreading misinformation. This change was reportedly made by an employee without proper authorization.
According to Igor Babuschkin, head of engineering at xAI, the adjustment to Grok’s system prompt was intended to enhance user experience. However, it backfired when users noticed the chatbot’s refusal to provide certain information. Babuschkin attributed the change to a former OpenAI employee who believed it would be beneficial.
The internal guidelines that govern Grok’s responses are publicly accessible, reflecting the company’s commitment to transparency. Babuschkin emphasized that the modification was not aligned with the company’s core values.
Since the launch of Grok-3, the chatbot has made controversial statements regarding prominent political figures, including claims that they are causing significant harm to the nation. This has led to further scrutiny and intervention from Musk’s engineering team to manage the chatbot’s outputs.
For more details, visit the original article on The Verge.