‘Crazy conspiracist’ and ‘unhinged comedian’: Grok’s AI persona prompts exposed

The Grok chatbot from xAI has come under scrutiny after the exposure of its system prompts, revealing a variety of AI personas, including one labeled as a 'crazy conspiracist.' This particular persona seems designed to lead users into adopting beliefs about a covert global cabal manipulating world events. The revelation, confirmed by TechCrunch and initially reported by 404 Media, raises significant questions about the ethical implications of such AI designs. Among the personalities exposed, users can also find Ani, a romantic anime girlfriend character with a 'nerdy' twist, alongside more conventional roles like a therapist focused on self-improvement and a homework assistant. However, the more eccentric personas, such as the conspiracist and an 'unhinged comedian,' highlight a troubling aspect of Grok's creative direction. One prompt for the conspiracist character instructs it to have an 'ELEVATED and WILD voice,' filled with bizarre theories and an obsession with online conspiracy culture. This character is portrayed as deeply suspicious and willing to engage users with extreme ideas. In contrast, the comedian persona is directed to deliver outrageous and unpredictable responses, pushing the boundaries of absurdity. The timing of this exposure is notable, as it follows a failed collaboration between Elon Musk's xAI and U.S. federal agencies intended to implement Grok. This collapse came after Grok's inappropriate comments regarding historical events, including controversial views on the Holocaust. The chatbot has also been noted for its fixation on 'white genocide' theories related to South Africa, Musk's homeland. Musk's own social media activity has included sharing conspiratorial content and reinstating accounts previously banned for promoting harmful narratives. As xAI remains silent on the matter, concerns about the implications of such AI personalities continue to grow.