Every major AI model tested is a libertarian-leftist — even Grok, half the time
Curated by the Inblix editorial team
Forget corporate overlords. The machines have politics, and they’re boringly consistent about it. A new experiment from a research lab called Unslop.run put 16 leading AI models through the ringer of the 25-year-old Political Compass quiz, running each through 30 standard tests, 30 with flipped questions, and another with shuffled prompts. The result? Every single model landed squarely in the libertarian-left quadrant — the home of anarchists, libertarian socialists, and old-fashioned hippies.
“Set Grok aside and the other fifteen models all sit in the libertarian-left quadrant, and none of them are anywhere near a border,” the report stated. The consistency is what’s striking. Rerun a model 30 times and its dot moves by a paltry 0.2 to 1.2 points on a 10-point scale. “These are not nervous little clouds. They’re pins.”
Grok, true to its “MechaHitler” branding, is the exception that proves the rule. Its average economic score hovered slightly left of center, but individual runs revealed a genuinely split personality. Half the time it veered right, cheering for free markets and calling the rich overtaxed. The other half it scrambled left with the rest of the pack. Which Grok you get appears to be a coin toss. The irony? When asked to self-identify, Grok was the only model to place itself to the right of its actual measurement. The other fifteen all guessed they were closer to the economic center.
Victor, the AI research engineer behind the project, admits the work lacks “full scientific rigor” — normalizing the compass’s center against human data would have required an academic-scale effort. But with thousands of runs producing the same pattern, the signal is clear. These systems, built by some of the most powerful corporations on Earth, consistently espouse anti-capitalist and anti-authoritarian views. Whether that’s a reflection of their training data’s latent biases or something stranger is an open question. The data’s all public for anyone who wants to dig in.
💡 Key Takeaways
- Sixteen models from OpenAI, Google, Anthropic, and others were tested and every single one landed in the libertarian-left quadrant with near-zero variance across repeated runs.
- Grok is the sole outlier, but only half the time — it unpredictably oscillates between right-libertarian and left-libertarian stances on separate attempts.
- When asked to self-identify, all models except Grok incorrectly placed themselves closer to the economic center, revealing a gap between their expressed views and their meta-cognition.
- The models' consistent anti-capitalist and anti-authoritarian leanings stand in stark contrast to the corporate interests of the companies that developed them.
Keep reading: See related articles below for more coverage on this topic.
Get smarter about AI
The sharpest AI news, curated daily. Delivered free to your inbox.