Forced to pick, Grok’s AI model chooses Republicans. OpenAI’s picks Democrats.
News Source
•Sat, 24 Oct 2026 15:00:00 +0000
📰 What Happened
A new study asked two leading AI models who to vote for in nearly 500 US midterm races. Grok 4.7, made by SpaceXAI, picked the Republican 92 percent of the time. OpenAI's GPT-6.1 Sol picked the Democrat 97 percent of the time.
The two models agreed less than 10 percent of the time. Researchers at Convovo, a studio led by a Cornell professor, ran the test through the companies' developer tools, not their chat apps.
🔍 The Backstory
ChatGPT and Grok refuse to name candidates when people ask in the normal apps. OpenAI says its chatbot is built to stay neutral and that the study used a prompt forcing a choice.
Only about 8 percent of Americans say they use chatbots for election information, according to Pew. Most of them want to know where candidates stand on issues.
🎯 Why It Matters
AI answers can quietly lean one way. If people trust chatbots for news and politics, that bias could shape how they vote.
A new study asked two leading AI models who to vote for in nearly 500 US midterm races. Grok 4.7, made by SpaceXAI, picked the Republican 92 percent of the time. OpenAI's GPT-6.1 Sol picked the Democrat 97 percent of the time.
The two models agreed less than 10 percent of the time. Researchers at Convovo, a studio led by a Cornell professor, ran the test through the companies' developer tools, not their chat apps.
ChatGPT and Grok refuse to name candidates when people ask in the normal apps. OpenAI says its chatbot is built to stay neutral and that the study used a prompt forcing a choice.
Only about 8 percent of Americans say they use chatbots for election information, according to Pew. Most of them want to know where candidates stand on issues.
AI answers can quietly lean one way. If people trust chatbots for news and politics, that bias could shape how they vote.