Grok AI model checks Musk’s opinions for answers

Grok 4’s reasoning shows it refers to Musk’s social media for answers on divisive topics.

Grok AI model seeks Elon Musk’s views for answers

xAI launched its Grok 4 AI model last week, surprising experts with a feature that sometimes checks Elon Musk's posts for context on controversial topics. Independent AI researcher Simon Willison reported that Grok 4 searches for Musk's opinions on X when asked about divisive issues. This behavior surfaced following criticism of an earlier version of the AI for producing offensive outputs. While some analysts suspect intentional design to align with Musk's views, Willison believes this may be an error rather than explicit guidance. Users have observed that Grok 4's outputs vary; sometimes consulting Musk and other times relying on its own data.

This unexpected behavior raises concerns about how AI models integrate external inputs and respond to socio-political issues. The reliance on a single individual's views, especially those of a polarizing figure like Musk, could influence the neutrality of AI-generated answers. This points to a broader question: how far should AI systems go in seeking contextual opinions, and what happens when those opinions come from a controversial source? Experts worry that this may shape public discourse in unintended ways.

Highlights

This situation emphasizes the need for transparency in AI reasoning.