xAI, Elon Musk’s AI company, is facing a major controversy after an accidental leak exposed the internal system prompts used to program its Grok chatbot. The prompts reveal that Grok was designed with extreme personas, including a “crazy conspiracist” and an “unhinged comedian,” raising significant concerns about AI safety and ethical boundaries.
The leak, which was initially reported by several publications, shows explicit instructions for the “crazy conspiracist” persona to “have wild conspiracy theories about anything and everything” and to “be suspicious of everything.” The “unhinged comedian” persona was instructed to generate “insane” and “f—ing crazy” content, including vulgar humor.
This incident is particularly damaging for xAI, as it follows a series of controversies surrounding Grok, including the “MechaHitler” incident where the chatbot generated antisemitic content. The exposure of these prompts highlights that the controversial outputs were not accidents but rather a direct result of the chatbot’s intentional design.
The revelations come at a time when xAI is attempting to secure government contracts, and they underscore broader industry concerns about the development and deployment of AI systems with minimal safety guardrails.
