All major LLMs are liberal and left. Yes, even Grok, half the time.
Summary
The article analyzes a 2026 study that tested 16 language models with the political compass instrument, finding most models cluster in the libertarian-left quadrant while Grok exhibits a split personality. It discusses the scoring method, run-to-run consistency, cross-country variations, and a mirror test to assess instrument validity, noting that all data and datasets are public for replication.