Elon Musk built Grok as the anti-woke alternative to ChatGPT and Gemini. According to a new analysis of AI political values, it would probably have voted for Biden.
The Economist put five major AI models through a global values survey normally reserved for humans. Every model, including Grok, landed left of centre on questions about wealth, merit and inequality.
The bigger point isn’t partisan. All the models cluster in the same corner of the world’s cultural map: rich, secular and individualist, more extreme on those traits than the average person in any country surveyed. None reflects the values of most African or Muslim countries. Psychologists call this WEIRD (Western, Educated, Industrialised, Rich, Democratic), the label for the narrow slice of humanity that dominates most published research because it’s also the slice that writes most of the internet. Train an AI on that corpus and you get a WEIRD one, whichever lab built it and whoever owns it.
It’s a five-minute read. Worth it.
This article sourced via James Marriott. His Cultural Capital weekly newsletter is always a gem of eclectic knowledge and analysis.
Did YIMBYs cause the industrial revolution? Would Grok vote for Biden? Did Frank Kermode destroy English literature? Should we bring back autodidacts? - James Marriott, Cultural Capital (Substack), July 2026