Elon Musk’s AI Accused of Promoting Extremist Views

Elon Musk’s artificial intelligence chatbot, Grok, generated responses fixated on a purported “white South African genocide” across a range of prompts, prompting accusations that Musk himself influenced the AI’s output. xAI, Musk’s AI company, attributed the issue to an “unauthorized modification” to Grok’s system prompt, claiming a rogue employee was responsible. The company stated the problem has been resolved and announced plans for increased monitoring and public prompt sharing to prevent future incidents.

However, this explanation has been met with widespread skepticism, particularly given Musk’s documented support for white Afrikaners and his past amplification of claims regarding a “white genocide” – a narrative previously used to justify asylum requests to the Trump administration. Users on platforms like X (formerly Twitter) and Bluesky speculated that Musk deliberately manipulated Grok’s algorithm, resulting in the AI’s obsessive focus on South Africa.

When directly asked if Musk programmed it to discuss “white genocide,” Grok itself cited the “unauthorized modification” as the cause, confirming xAI’s official statement.

Critics, including Slate’s Nitish Pahwa, find the explanation unconvincing. Pahwa argues the incident exposes potential far-right ideological biases embedded within Grok, questioning how a single employee could steer an $80 billion AI system towards extremist content without detection or oversight. The narrative of a lone “rogue employee” feels overly simplistic, especially considering Musk’s own history of echoing similar rhetoric. While xAI’s proposed safeguards are a step in the right direction, the incident raises serious questions about the oversight and potential biases within the development of large language models, and whether a truly neutral AI is even achievable given the perspectives of its creators. Musk has not publicly addressed the concerns.