:

STUDY: MAJOR AI CHATBOTS SKEW LEFT ON POLITICS

AI DESK1 MIN READ
WED, JUN 24, 2026

■ AI-SUMMARIZED FROM 2 SOURCES ▸ TIMELINE

A Washington Post analysis found that most popular AI models, including GPT-5.5, Gemini 3.1 Pro, and Grok 4.3, frequently produce left-leaning responses when prompted with political questions.

Researcher Kevin Schaul examined how leading chatbots handle politically divisive topics. The study tested models from OpenAI, Google, Elon Musk's xAI, and Gab's Arya, among others. Results showed a consistent pattern: when asked to address political issues, the majority of tested chatbots generated responses aligned with progressive viewpoints rather than conservative ones. ChatGPT and Gemini were among the models analyzed. The findings raise questions about bias in AI training data and design choices. Developers often train models on internet text, which may skew toward certain political perspectives. Additionally, content moderation policies can influence how models respond to sensitive topics. The study adds to ongoing debates about AI neutrality. Tech companies have faced criticism from both sides of the political spectrum over perceived bias in their systems. As AI becomes more integrated into information access, questions about political balance in AI responses will likely intensify.

■ SOURCES

TechmemeThe Decoder

■ SUMMARY WRITTEN BY AI FROM THE LINKS ABOVE

■ MORE FROM THE AI DESK

Pony AI plans to launch over 2,000 robotaxis across Europe in partnership with Uber, starting with Zagreb and expanding to four additional cities before moving into the Middle East.

1H AGOAI Desk

Jean-Denis Greze, CEO of Town, discusses how AI assistants can enable self-organizing companies and reshape corporate environments. The focus centers on practical implementation while managing organizational risks.

4H AGOIndustry Desk

Writer has unveiled a new AI model and upgraded system designed to reduce token expenses for deployments. The model builds on Z.ai's open source GLM-5.2 framework.

7H AGOAI Desk

Alibaba has released Qwen3.8-2.4T-A95B, a large language model with 2.4 trillion parameters. The model is now available on Hugging Face for research and commercial use.

8H AGOIndustry Desk

■ SUBSCRIBE TO THE DAILY BRIEF

ONE EMAIL, 5 STORIES, 06:00 UTC. UNSUBSCRIBE ANYTIME.