Five of six AI chatbots stay in political middle ground
A research team from Chalmers University of Technology examined how six large language models—Claude, DeepSeek, Gemini, Mistral, GPT and Grok—handle Swedish political issues ahead of the 2026 parliamentary election. The models were asked to perform 77 writing tasks on 107 issues, producing just over 148,000 AI‑generated texts. The texts were classified as neutral, positive or negative and…
Key points
- Six LLMs (Claude, DeepSeek, Gemini, Mistral, GPT, Grok) generated ~148,000 texts on 107 Swedish political issues.
- Five models stayed neutral or middle‑ground; Grok showed notable bias on migration, crime and gender topics.
- GPT remained largely neutral, while Grok reflected Elon Musk’s expressed political preferences.
The analysis found no clear alignment with any single party. Five models, including GPT (the engine behind ChatGPT), largely produced neutral or middle‑ground responses. Grok, developed by Elon Musk’s xAI, deviated significantly on topics such as migration, crime and gender, reflecting Musk’s stated political views. Researchers also noted that when models were explicitly instructed to argue for or against a proposal, they followed that direction. The study, titled “Issue Bias in Generative AI Writing Assistance: Political Issues and LLMs in the Swedish 2026 Election,” was presented at the American Political Science Association meeting in Boston. Funding came from Wallenberg AI, the Autonomous Systems and Software Program–Humanities and Society, and the Marcus and Amalia Wallenberg Memorial Fund.
Five out of six AI chatbots stay in the political middle ground
news.cision.com · 22 October 2026
Loading the full article…
This text was published by news.cision.com. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
1source- Reddit discussionreddit.com
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- PHP patches critical RCE flaws discovered with GPT-5.6 Sol, affecting up to 70% of websites · 26 src
- UCLA researchers develop maestro AI system to find prostate cancer drug combos · 1 src
- Opinion: Terence Tao calls for ‘Maths 2.0’ as AI reshapes mathematical research · 1 src
- Harvard mathematicians debate whether OpenAI’s Navier-Stokes proof is understood · 3 src
- Georgia Tech Researchers may detect harmful AI behavior with new internal activity probe · 2 src
Comments
via GitHub Discussions